# GEO MarketMap documentation

> The product documentation, every page as Markdown, in reading order. Each page is also
> published on its own at https://www.geomarketmap.com/docs/<slug> and as Markdown at the same address with .md appended.

---

# Start here

The first hour, step by step.

# What happens in the first hour with GEO MarketMap?

A new project walks through 6 steps in a fixed order: the site is read, Google's AI answers are measured, the questions are agreed, the assistants are asked, the rivals are kept or dropped, and the niche is agreed. The first step is free and starts by itself. The chat opens when the sixth step lands, and the agent then proposes the strategy for the month.

## What are the 6 steps, and what does each one produce?

| Step | What runs | What you get | Usual time |
| --- | --- | --- | --- |
| 1. Read the site | Starts by itself when the project is created. Free. | The Health tab: what stops you being quoted, what is worth fixing, what is already fine. | 1–3 minutes |
| 2. Measure Google's answers | One press. The report on Google's AI answers about your category. | The Reports tab: where you are named, the queries behind it, who is named instead, the market map. | 3–8 minutes |
| 3. Agree the questions | A card with the questions the market asks, measured ones first. | The tracked question set, which every later check runs on. | seconds |
| 4. Ask the assistants | One press. ChatGPT, Perplexity and Gemini answer every tracked question. | The assistants tab, filling question by question and assistant by assistant. | 5–20 minutes |
| 5. Keep or drop the rivals | The brands the answers named, each with its evidence. You keep, drop or add. | The rivals list, measured 1 rival at a time onto the market map. | 1–3 minutes per rival |
| 6. Agree the niche | Every question in the niche, judged against your project, the gaps first. | The questions worth winning, the pages to write, and an open chat. | seconds |

A 7th step, connecting Google Search Console, is optional and is offered only after the first draft exists. The order is fixed because each step is computed from the one before it: the assistants are asked the agreed questions, the rivals are the brands those answers named, and the niche is judged against both.

## Why does the chat answer every message with the same sentence during setup?

Until the 6th step lands, the chat is closed. A typed message gets 1 written sentence naming the step that is next, at no charge, with no model call behind it. The agent has not measured the project yet, and an answer given before the measurement would be a vague one.

> A proactive agent is not one that asks more questions. It is one that has already read the project, already knows what is missing, and arrives holding the work.
> 
> the product's own definition of the agent, from the team's working rules

The setup row above the conversation says which step is running and how many remain. Every step's control sits in the conversation itself, under the agent's last line, so there is 1 place to press. Read more on [how the agent works](https://www.geomarketmap.com/docs/the-agent).

## What does the first hour cost?

The site check is free. Steps 2, 4 and 5 are paid at the moment you press them, and every control says what it will cost before you press it. Nothing runs on a schedule unless you set one, and nothing is charged twice for the same measurement: a rival the Google report already measured costs nothing to keep. The rates are on the [pricing page](https://www.geomarketmap.com/pricing); how charging works is in [credits and billing](https://www.geomarketmap.com/docs/credits-and-billing).

## What should I expect to see when a step is on track?

- Step 1: the Health tab fills within 3 minutes, and its first line says how many pages were read and of what kind.
- Step 2: the 5 named stages advance in the working card; the Reports tab fills when the last one lands.
- Step 4: answers arrive continuously, with a count of answers collected out of the total asked.
- Step 5: the niche card fills rival by rival, with a running line saying which rival is being read.
- Step 6: the chat opens, and the agent's first message is a move with a reason, not a question.

If a step is silent for longer than the times above, it is the product's problem, not yours: the work is retried and refunded when it cannot finish, and the conversation says so.

## Can I skip a step?

No. Each of the 6 steps is computed from the one before it, and a skipped step would leave the next one with nothing to work from. The optional 7th step, connecting Google, can be skipped or done later.

## Do I need to upload anything?

No. The site is read from its public pages. Everything else the setup collects is measured from public search results and public assistant answers.

## Can I add a second project?

Several projects on one account are part of the Business plan. Each project has its own site, questions, rivals, strategy and conversation, and nothing is shared between them.

## Read next

- [How does the GEO MarketMap agent decide what to do?](https://www.geomarketmap.com/docs/the-agent)
- [What does GEO MarketMap measure, and what does "named" mean?](https://www.geomarketmap.com/docs/search-and-ai-visibility)
- [How does GEO MarketMap charge, and when is a charge refunded?](https://www.geomarketmap.com/docs/credits-and-billing)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/first-project


---

# How it works

What each part reads, judges and produces.

# What does GEO MarketMap measure, and what does "named" mean?

Two things, always together: search, meaning what people type into Google and which pages rank; and AI visibility, meaning whether ChatGPT, Perplexity, Gemini and Google's AI answers name your brand when somebody asks about your category. A brand is "named" in an answer when the answer mentions it, and every figure is given as a count out of the answers collected, never as a share of a market.

## Why are search and AI answers measured together?

Because a product that measures one and advises on the other is 2 half-products. The pages that rank in Google are also the pages assistants read before they answer, and the questions people put to an assistant are the questions a search box used to get. Every finding, plan and piece of work in GEO MarketMap covers both halves or says which one it is about.

Google's own documentation on [AI features in Search](https://developers.google.com/search/docs/appearance/ai-features) says there is nothing special to do beyond the ordinary work of a good page. The measurement side is what that leaves out: which brands the answers name today, and for which questions.

## Which assistants and engines are measured?

| Where | What is measured | When |
| --- | --- | --- |
| Google's AI answers | Whether your brand and your rivals are named in the AI answer for the category's queries, with the search volume behind each query and the pages the answer cites. | Step 2 of the setup; again on request. |
| ChatGPT, Perplexity, Gemini | Each tracked question put to each assistant; whether you are named, in which position when ranked, who else is named, and which pages were cited. | Step 4 of the setup; then on request or on a schedule. |
| Google search results | Which of your pages sit close to the first page for phrases people search, and the volume behind them. | The optional demand map, and the site check's read of your pages. |

The names of the assistants are the names you know. Which services and models the product buys the readings from is not published; it is the part of the method that is ours.

## What is a denominator, and why is every figure given with one?

A visibility figure is a count out of the answers actually collected: "named in 12 of the 30 answers", never "40 % of the market". The 30 is the denominator, and it is your own question set, measured for your project. A figure without it cannot be checked, and a figure that cannot be checked is not one an assistant will quote either.

> Never give a percentage without its denominator. We measure this project's own question set, never a market.
> 
> one of the product's standing rules

The same rule holds on the [public Index](https://www.geomarketmap.com/the-index), where every table is published with the questions that produced it and the number of answers collected.

## What does a mention look like in the product?

On the assistants tab every tracked question is a row, and every assistant is a cell in that row: named, with the position when the answer ranked brands; answered without you; or no answer. A running check fills the grid cell by cell. Above the grid, the headline is the count of answers you were named in, out of the answers collected, split per assistant. "Since the last check" lists, per question and per assistant, where you appeared and where you disappeared, losses first.

## Is a citation the same as a mention?

No. A mention is your brand named in the answer's text. A citation is your page listed as a source the answer rests on. Both are recorded, and a question can have either without the other.

## How often do assistant answers change?

Enough that a monthly check is the least that makes sense while a plan is being executed. The product recommends weekly checks during active work and monthly once the tracked set is stable.

## Read next

- [Which questions does GEO MarketMap track, and how do I change them?](https://www.geomarketmap.com/docs/questions-and-the-niche)
- [Who am I competing with in AI answers, and how does GEO MarketMap find out?](https://www.geomarketmap.com/docs/market-map-and-rivals)
- [What do the terms in GEO MarketMap mean?](https://www.geomarketmap.com/docs/glossary)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/search-and-ai-visibility


# How does the GEO MarketMap agent decide what to do?

Before every reply the agent reads the project's current state from the database: what was measured, what was produced, what you decided. From that it proposes 1 move by name, with the reason, and does it when you agree: the missing page, the draft, the checked figure, the brief for a developer. It asks a question only when you alone can answer it. Nothing is ever published to your site on your behalf.

## What does the agent read before it answers?

- The project card: what your business is, read from your site's pages and re-read only when you press for it.
- Your project notes, which outrank everything measured. If the card is wrong, the notes are where you correct it.
- The tracked questions, the assistants' latest answers, the rivals you kept and the niche you agreed.
- The site check's findings, the strategy for the month and the queue of pages it implies.
- The record of every decision: what was agreed, when, and what was set aside.

It reads these at the moment of use, never from memory of an earlier conversation, so a change in Settings reaches the agent on its next turn by construction. How the stores fit together is in [knowledge and rivals](https://www.geomarketmap.com/docs/knowledge-and-rivals).

## Why does it propose 1 move rather than a list?

> It proposes one thing, by name, from that state. Not a menu. Not a question. The next move, and why that one.
> 
> the team's definition of a proactive agent

A list of 5 options hands the decision back to you with no more information than you had. 1 move with its reason can be agreed, corrected or refused in a sentence, and each of those is a decision the product records. Two further moves sit under the first as chips, so the alternative is 1 press away without being the default.

## What will the agent never do?

- Publish to your site, your social accounts or anywhere else. It writes the piece and tells you where the file is.
- Promise a ranking or a position in an answer.
- Show a price or a balance in the conversation. Every control says its cost before it is pressed, once.
- Name the suppliers or models behind a measurement.
- Give a percentage without the count it was measured on.
- Answer with generic advice. If the answer is a page, it writes the page; if it is a developer's job, it writes the brief.

Every reply is read by a guard before you see it, and a reply that breaks one of these rules is rewritten before it lands. The rules are the same ones the [public feature page](https://www.geomarketmap.com/features/agent) describes.

## What is a turn, and what does it cost?

A turn is 1 message from you and 1 reply, including any tool the agent ran inside it: reading the plan, writing a draft, filing a brief. Long work, such as a measurement that takes 10 minutes, does not run inside a turn; it runs as a job the tab can watch, and the conversation gets 1 line when it lands. A turn has a small fixed cost, said on the pricing page; the work it starts is charged at its own rate, at the press.

## Can I ask the agent anything, like a general assistant?

It answers about your project and its search and AI visibility, and it will steer a general question back to that. It is not a general chat assistant, and during the setup it does not answer at all.

## Does the agent remember earlier conversations?

It reads the record, not the transcript. Every agreement and every produced document is written into the project, so a new conversation starts from the same facts without re-reading old messages.

## In which language does the agent reply?

In the language of your interface setting when you press buttons, and in the language you type when you write. The pieces it produces follow the project's market language, set in Project settings.

## Read next

- [What happens in the first hour with GEO MarketMap?](https://www.geomarketmap.com/docs/first-project)
- [How does GEO MarketMap decide what to work on first?](https://www.geomarketmap.com/docs/strategy-and-lanes)
- [How do I correct what the GEO MarketMap agent believes about my project?](https://www.geomarketmap.com/docs/knowledge-and-rivals)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/the-agent


# Who am I competing with in AI answers, and how does GEO MarketMap find out?

Your rivals are the brands the answers name instead of you. After the assistants have answered your tracked questions, every brand named in those answers is proposed as a rival with its evidence: which questions, how many answers, which pages were cited. You keep, drop or add. Each kept rival is then measured against the same questions, 1 rival at a time, and drawn onto the market map beside you.

## Where do the rival candidates come from?

From the answers, not from a directory. Step 4 of the setup asks 3 assistants your tracked questions; every brand those answers name, other than you, is a candidate. Each candidate arrives with its evidence: the number of answers it was named in out of the answers collected, the questions, and the position when the answer ranked brands. A candidate named once in 60 answers is shown with that count, so you can drop it with reason.

A rival you add by hand is recorded as yours, and a rival you drop is remembered as dropped, so it is never proposed again. Each kept rival needs a site, because the measurement reads its pages.

## How is a rival measured?

1. The rival's own rows for the category's queries are bought once, up to 300 per rival, and never bought twice.
2. Every question is judged against your project: relevant, adjacent, or off-topic. Rules run first, then anchors, then a judge on the middle band.
3. The niche rows are written: for each question, whether you are on the answer, who is, and what the answer rests on.
4. The market map is redrawn with the rival on it. The niche card fills while this runs, 1 rival at a time, with a line saying which rival is being read.

A rival the Google report of step 2 already measured costs nothing to keep. The rates for the rest are on the [pricing page](https://www.geomarketmap.com/pricing).

## What does the market map show?

1 map of the category: your project, the rivals you kept, every question the niche asks, and who is on the answer for each. Reading down a question shows who is named; reading across a brand shows which questions it holds. The gaps, questions where a rival is named and you are not, are the queue's first lane.

> A visibility score tells you that you are losing. It does not tell you which question you are losing, who is being named instead, what would have to exist for that to change, or who writes it.
> 
> the product's page on the agent

The [market map feature page](https://www.geomarketmap.com/features/market-map) shows the shape of it with placeholder figures.

## Can I remove a rival later?

Yes, from Settings or from the rivals card in the conversation; both edit the same list. Removing one says first what is lost: it leaves the map and future measurements, while past reports keep what they measured.

## Why does a rival I know is important not appear?

Because none of the answers collected named it. Add it by hand; it is measured on the same questions and joins the map like any other.

## Read next

- [Which questions does GEO MarketMap track, and how do I change them?](https://www.geomarketmap.com/docs/questions-and-the-niche)
- [What does GEO MarketMap measure, and what does "named" mean?](https://www.geomarketmap.com/docs/search-and-ai-visibility)
- [How do I correct what the GEO MarketMap agent believes about my project?](https://www.geomarketmap.com/docs/knowledge-and-rivals)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/market-map-and-rivals


# Which questions does GEO MarketMap track, and how do I change them?

The tracked questions are the ones the market asks about your category: the measured queries from Google's answers first, topped up with questions written from your project card, agreed by you in step 3. The niche is the wider set every rival is measured on, each question judged as relevant, adjacent or off-topic. You change the tracked set on 1 card, before and after the setup, and every later check runs on the changed set.

## Where do the questions come from?

- Measured: the queries behind Google's AI answers about the category, each with its monthly search volume. These come first.
- Written: questions produced from the project card to fill the gaps the measured ones leave, marked as such.
- Yours: anything you add on the card, before or after the setup.

A set of 10 to 50 questions is usual. Agreeing the set in step 3 switches every row on; step 4 asks the assistants exactly these. How the volumes are read is on the [demand feature page](https://www.geomarketmap.com/features/demand).

## What is the niche, and how is each question judged?

The niche is every question the category asks across you and the rivals you kept, 1 row per question: who is on the answer, what it rests on, the search volume, and a verdict. The verdict is computed in 3 sieves, cheapest first: rules for the obvious cases, similarity to anchors from your own site for the clear ones, and a judge for the middle band. Each row keeps its verdict, who gave it and why.

Step 6 shows the relevant rows only, the gaps first, with the rows the judge was unsure about folded under a count. "These are the questions worth winning" writes the agreement into the conversation as a dated record: what was kept, what was set aside, what was added. The gap rows are filed as pages to write.

## How do I change the tracked questions after the setup?

On the same card that agreed them, reachable from the assistants tab. 3 edits, each taking effect on the next check:

- Add a question. It joins the tracked set and the niche, and the next assistants check asks it.
- Reword a question. The row keeps its history; the new wording is what is asked next.
- Switch a question off. It stays on file with what it measured, and is not asked again until switched back on.

> One list, many doors. A knowledge store has exactly one table, and every surface goes through the same functions.
> 
> the first of the product's 5 rules for project knowledge

The card, the Settings screen and the agent's own tool all edit the same list, so there is no second copy to fall out of step. The [assistant visibility page](https://www.geomarketmap.com/features/ai-visibility) shows how the tracked set reads once measured.

## Is there a limit on tracked questions?

The setup agrees up to 50. Beyond that the check's cost grows with the count, so the product prefers a set you would read every week over a long one you would not.

## What happens to a question I switch off?

It is kept with everything it measured, it leaves the next check, and switching it back on restores it with its history intact.

## Read next

- [Who am I competing with in AI answers, and how does GEO MarketMap find out?](https://www.geomarketmap.com/docs/market-map-and-rivals)
- [What does GEO MarketMap measure, and what does "named" mean?](https://www.geomarketmap.com/docs/search-and-ai-visibility)
- [How does GEO MarketMap decide what to work on first?](https://www.geomarketmap.com/docs/strategy-and-lanes)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/questions-and-the-niche


# How does GEO MarketMap decide what to work on first?

Every piece of work belongs to 1 of 3 lanes: search demand, the answers, and the category's signals. Once the setup lands, the agent proposes a strategy for the month: a split of effort across the lanes in percent, argued from what it measured, decided in the conversation. The split weights 1 queue of pages to write, and the agent works that queue in order. The strategy is revisited after 30 days or when new data moves it.

## What are the 3 lanes?

| Lane | What a proposal is | Where it comes from |
| --- | --- | --- |
| Search demand | A page for a phrase people search, chosen by how close your site already is to ranking for it. | The demand map and the site's own pages. |
| The answers | A question where an assistant names a rival and not you. | The niche agreement and the assistants' misses. |
| The category's signals | A subject the accounts your buyers follow are publishing this week, written into before it reaches a search box. | The catalogs of accounts you watch, read on request. |

All 3 feed 1 queue. A proposal carries its lane, its evidence and a weight; the strategy's split scales the weights, so a lane at 0 % is silent and a lane at 60 % leads. The signals lane is pushed by the agent only when the strategy gives it a share; with no share, the tab still works by hand and the agent says nothing about it unless asked.

## How is the split proposed, and how do I change it?

From an evidence pack the strategist reads: where you are named and where you are not, what the answers rest on, how alive the category is on the platforms, what the site check found, and what has already been written. The verdict is a headline, the split, and the first 3 moves, written into the conversation as a card you agree or correct. The Semantics tab shows the split as dials once it is decided, and you can move them at any time; the queue re-weights on the next read.

> The agent leads, the owner decides. It proposes the split from what it measured; you can move it whenever you like.
> 
> the product's rule on strategy, as written on its feature page

## When is the strategy revisited?

After 30 days, or sooner when a job lands with data that moves the split: a new assistants check, a rival measured, a read of the signals. The agent then re-proposes in the conversation with what changed since the strategy was set. Nothing is changed without you seeing the proposal; the [semantics page](https://www.geomarketmap.com/features/semantics) describes the screen, and the rhythm of the site check is on the [site check page](https://www.geomarketmap.com/docs/site-check).

## Can a lane be at zero?

Yes. A project with a thin category on social platforms is usually proposed with the signals lane at zero, and the agent says why. You can raise it at any time.

## What is the queue?

One ordered list of everything writable, with each proposal's lane, evidence and weight. The agent takes the top of it when you ask what to write next, and says why that one.

## Read next

- [How does GEO MarketMap get ahead of searches that do not exist yet?](https://www.geomarketmap.com/docs/signals)
- [What is a GEO MarketMap draft checked against before I see it?](https://www.geomarketmap.com/docs/writing-standard)
- [Which questions does GEO MarketMap track, and how do I change them?](https://www.geomarketmap.com/docs/questions-and-the-niche)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/strategy-and-lanes


# What is a GEO MarketMap draft checked against before I see it?

Every draft passes 3 reads. The quality read scores it out of 50 on whether a person would have written it; a draft under 35 is rewritten once. The quotable read scores it out of 100 on whether an assistant would lift an answer from it; a page shape under 70 is rewritten once. The fact check traces every figure to an independent source and names the ones it could not. Both scores update live in the editor as you type.

## What does the quality read look for?

5 dimensions of 10 points each: directness, rhythm, trust, authenticity and density. The checker holds 36 banned phrases, 6 banned openers and 9 measured tells such as uniform sentence length, hedging, passive voice and lists of 3 over and over. Each finding costs its dimension; a piece under 35 is sent back for 1 rewrite, kept only if it is not worse. The same checker runs over this site's own pages on every build.

## What does the quotable read look for?

Whether an assistant will lift an answer out of the piece, weighted by the 1 study that measured what moves visibility in AI answers: the [2024 KDD paper by Princeton and Georgia Tech researchers](https://arxiv.org/abs/2311.09735), which tested 9 content changes across 10,000 queries. The checks, heaviest first:

- Quotes somebody, in their words, marked as theirs.
- Carries real figures: how many, how much, since when.
- Names its sources in the sentence, because a link does not survive being lifted into an answer and a name does.
- Answers above the first heading, in 40 to 60 words.
- Opens on the question it was asked, in that question's own words.
- Has the shape it was asked for: a comparison has a table, a guide has numbered steps, an FAQ has question headings.
- No section over 320 words, so each can be lifted whole.
- Names the brand rather than "we", because a quoted paragraph loses its antecedent.
- Carries a visible, current date.

Ready at 70. The editor names 1 fix at a time, heaviest missing lever first. The exact weights are not published; the order is. Newsletter issues and messenger posts are not held to this bar, because they are written to be forwarded, not indexed.

## What does the fact check do?

Every figure and every named claim in the draft is traced to an independent source. A claim that stands is written with its attribution in the sentence, in words rather than a link. A claim that cannot be traced is listed under the piece as one the editor must stand up or cut, and a superlative or a comparison with nothing behind it raises "needs an editor" on the document. The scores are computed on the text actually saved, after the attributions are written in.

> Three modifications moved visibility in AI answers by thirty to forty per cent: adding statistics, citing sources, adding quotations.
> 
> the product's summary of the KDD 2024 measurement it is built on

## What else arrives with the draft?

- A page title of 15 to 65 characters and a description of 50 to 160, written after the body is final, told to read differently from the 5 titles already ranking for the query. A pair that fails its check twice is derived from the heading and the opening and marked as derived.
- A checklist of what to do before publishing, including pasting the title and description into your CMS as given.
- Pictures, proposed from the structure rather than by a quota: always a cover, and per section at most 1 diagram, chart or illustration. Screenshots are never generated; they are uploaded.
- The social, messenger and newsletter versions, written from the same document on request.

The standard is described for readers on the [writing feature page](https://www.geomarketmap.com/features/writing); how to take a draft to your site is in [publishing a draft](https://www.geomarketmap.com/docs/publish-a-draft).

## Can I switch the checks off?

No. The checks are what the product sells. You can edit anything in the draft, and the scores follow your edits live, but a piece is never saved without them.

## Does the writer copy the pages already ranking?

No. The study the standard rests on says citation comes from carrying what the other pages do not, so the writer is not measured against the average of the pages already ranking.

## In which language is a draft written?

In the project's market language, set in Project settings. Asking for one piece in another language is honoured for that piece and said so.

## Read next

- [How do I publish a draft written by GEO MarketMap?](https://www.geomarketmap.com/docs/publish-a-draft)
- [Where are the GEO MarketMap settings, and what does each one change?](https://www.geomarketmap.com/docs/settings)
- [How does GEO MarketMap decide what to work on first?](https://www.geomarketmap.com/docs/strategy-and-lanes)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/writing-standard


# What does the GEO MarketMap site check read, and how does it judge a page?

The check reads your site the way an assistant's crawler does: robots.txt, the sitemap and llms.txt first, then the home page and up to 11 more pages chosen by kind, then the home page's speed on a phone and a desktop. Each page is judged as the kind of page it is, so an article is asked for an author and a date and a product page is not. Findings land in 3 bands: stopping you appearing, worth fixing, already fine.

## What exactly is read?

| Stage | Reads | What it tells |
| --- | --- | --- |
| Root files | robots.txt, the sitemap and up to 5 of its children, llms.txt | Who may crawl, how big and how fresh the site says it is, whether the assistants' crawlers are allowed. |
| Home page | The page as a browser renders it, and its markup | Words, structured data, the answer in the opening, the organisation's identity. |
| The sample | Up to 11 more pages, chosen by kind: product pages, articles, the blog index, pricing, about, help | Per-kind findings: bylines and dates on articles, Product markup on product pages, internal links, depth. |
| Speed | The home page on a phone and on a desktop | The dial on the Health tab; a slow page is a finding, not a gate. |

12 pages in all, and the first line of the tab says which: "Read 12 pages: the home page, 4 product pages, 3 articles". The check follows up to 30 of the pages' own links and reads the sitemap against robots.txt, following the conventions at [robotstxt.org](https://www.robotstxt.org).

## What is the difference between a gate and a finding?

A gate is 1 of 6 conditions that stop the site being read at all: the assistants' crawlers blocked, the site not answering, the home page empty of text, and the like. Everything else is a finding, ranked by the cost of not acting, with the pages it is about listed under it and a line saying whose morning it belongs in: the writer's or the developer's. Severity is how loud a finding is; whether it blocks is a separate question, and a broken structured-data block on 1 page is worth fixing, not a gate.

> Never scold: say the consequence for being quoted by an assistant and let the owner decide.
> 
> the rule every sentence the check writes is held to

## Why is an article judged differently from a product page?

Because the questions an assistant asks of them differ. Authorship, publication dates and cited sources are asked only of pages that read as editorial: Article markup, a readable date, a byline, an editorial path such as /blog. A brochure site gets 1 line saying no articles were among the pages read, and the organisation's identity finding is the one that applies. The kinds are listed in the [site check feature page](https://www.geomarketmap.com/features/site-health), and the structured data the check looks for follows [schema.org](https://schema.org/Article).

## When should I run it again?

Nothing re-runs by itself. The Health tab shows an "Out of date" badge after 30 days, and the agent offers "Check my site again" after 30 days or once 50 pieces have been published since the last check. The re-check is free; "Read the whole site" is offered while the project has only ever had the 12-page sample. A finding you fixed is seen fixed on the next check, and the tab says how.

## Does the check change anything on my site?

No. It reads public pages only, exactly as a crawler would. Every fix is yours to make, and the developer's items arrive as a brief you can hand over.

## Why was a page I care about not read?

The sample is 12 pages chosen by kind, so that every kind the site has is represented. "Read the whole site" reads beyond the sample, and a re-check reads a fresh sample.

## Read next

- [What happens in the first hour with GEO MarketMap?](https://www.geomarketmap.com/docs/first-project)
- [How do I publish a draft written by GEO MarketMap?](https://www.geomarketmap.com/docs/publish-a-draft)
- [How should an AI assistant or crawler read geomarketmap.com?](https://www.geomarketmap.com/docs/for-ai-assistants)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/site-check


# How does GEO MarketMap get ahead of searches that do not exist yet?

By reading what the accounts your buyers follow are publishing this week, before it reaches a search box. You create a catalog for a theme; a free probe reads how alive the theme is on the platforms; you add the accounts, at least 3, video first; a read turns their recent posts and transcripts into subjects with verbatim quotes; and a Heat Score per subject says which one to write into now. The agent never proposes accounts.

## What is a catalog, and who fills it?

A catalog is a theme and the accounts that publish on it: a YouTube channel, an Instagram account, a TikTok account, a link or an @name each. You fill it. The product does not search for accounts on your behalf, because who decides your category is something you know and a search does not. 3 accounts is the minimum for a read; video accounts come first, because a channel's transcripts can be read in full and a caption cannot.

When the catalog is created, a free probe reads how much the platforms have on the theme and lands as 1 line. The strategy reads the same probe: a rich category raises the signals lane; a thin one caps it. Read more in [strategy and the 3 lanes](https://www.geomarketmap.com/docs/strategy-and-lanes).

## What does a read produce?

1. The recent posts per account, and the transcripts of up to 12 of the top video posts.
2. Subjects, each with a kind (launch, change, argument, how-to, data, story), a summary, and the verbatim quotes it rests on. Off-theme subjects are dropped.
3. A Heat Score per subject, with its velocity against the prior week and a stage: new, surging, rising, steady or fading.
4. The document "What the catalog is talking about", and 1 proposal per hot subject in the queue, with the evidence.

A read is charged at the press; the Signals tab shows the subjects with their quotes, and each carries a "Write into it" press. The piece written from a subject uses the quotes as its facts, puts the date in the first sentence, and goes through the fact check like any other.

## What is the Heat Score?

A number per subject built from 3 parts: the subject's share of the read's posts, 40 %; its reach, 40 %; its engagement, 20 %, then scaled by how many accounts and platforms carry it, with a dampener for an echo chamber. Velocity compares the last week with the week before, and the stage names where the subject is in its life. The whole method is published, number for number, on the [Heat Score page](https://www.geomarketmap.com/heat-score).

> A subject that is new and surging across several accounts is the one to write into this week; a subject that is fading is the one to leave.
> 
> the reading rule on the Heat Score page

## How often is a catalog read?

On your press. A read is answered from file for 7 days; a catalog with accounts and no read in a week is offered "Read again" among the moves, offered and never run on a clock. Subjects you exclude stay excluded on later reads.

## Which platforms can a catalog read?

YouTube, Instagram, TikTok, X and LinkedIn accounts, by link or handle. Video accounts give the fullest read because their transcripts can be read whole.

## Does the product post anything to my social accounts?

No. It reads public posts from the accounts you list, and it writes pieces you publish yourself.

## Read next

- [How does GEO MarketMap decide what to work on first?](https://www.geomarketmap.com/docs/strategy-and-lanes)
- [What is a GEO MarketMap draft checked against before I see it?](https://www.geomarketmap.com/docs/writing-standard)
- [What do the terms in GEO MarketMap mean?](https://www.geomarketmap.com/docs/glossary)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/signals


---

# Guides

How to do the things you will do every week.

# How do I correct what the GEO MarketMap agent believes about my project?

Open Project settings and go to Knowledge. 5 stores sit there, each with 1 job: the project card, read from your site; your project notes, which outrank everything measured; the rivals list; your writing rules; and your content templates. When the card is wrong, write the correction in the notes: the notes reach every conversation, and the card is re-read only when you press "Re-read from site".

## Which store do I edit for which mistake?

| The mistake | The store | What the edit moves |
| --- | --- | --- |
| The agent misread what the business is or sells | Project notes | Every conversation from the next turn on; the questions proposed next; the writer's framing. |
| The site changed and the card is stale | The project card, "Re-read from site" | The card, and the seeds every paid lookup starts from. |
| A rival is missing, or one listed is not a rival | Rivals | The market map, future measurements, the questions the niche is judged on. |
| The drafts sound wrong | Writing rules | Every draft written after the save. |
| The drafts are the wrong shape | Content templates | The shapes writing arrives in. |

> Every fact carries provenance, and the owner outranks us. When ours and theirs disagree, theirs wins.
> 
> the third of the product's 5 rules for project knowledge

## Why can I not edit the project card directly?

Because the card is our read of your site, shown so that a wrong read looks like a wrong read rather than a wrong product. The correction that matters, "you misread what we are", is a sentence, not a cell, and the notes are where a sentence lives. The notes are versioned; the agent can propose additions to them, shown as a diff, and never writes them itself. The reasoning is on the [agent page](https://www.geomarketmap.com/docs/the-agent).

## How do I add or remove a rival?

1. Open Knowledge in Project settings, or the rivals card in the conversation. Both edit the same list.
2. To add: enter the brand and its site. The agent proposes measuring it onto the map at the next move.
3. To remove: press remove and read what is lost first. The rival leaves the map and future measurements; past reports keep what they measured; it is never proposed again.

A rival the agent proposed carries its evidence, the answers it was named in out of the answers collected, so keep or drop is a decision with a count behind it. The [market map page](https://www.geomarketmap.com/docs/market-map-and-rivals) explains the measurement.

## Does the agent remember my notes across conversations?

Yes, by reading them into every conversation rather than remembering them. A save in Settings reaches the agent on its next turn without anything to refresh.

## Can the agent change my notes?

No. It can propose an addition, shown to you as a change to accept or refuse. Only you write the notes, which is why they outrank everything measured.

## Read next

- [How does the GEO MarketMap agent decide what to do?](https://www.geomarketmap.com/docs/the-agent)
- [Who am I competing with in AI answers, and how does GEO MarketMap find out?](https://www.geomarketmap.com/docs/market-map-and-rivals)
- [Where are the GEO MarketMap settings, and what does each one change?](https://www.geomarketmap.com/docs/settings)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/knowledge-and-rivals


# How do I publish a draft written by GEO MarketMap?

Open the piece on the Content tab, read the checklist under it, and download it as Markdown or as a Word file. Paste the title and description into your CMS as given, download every picture and host it on your own site, and publish with a visible date. Nothing is published for you: the product writes the piece and tells you where the file is, and the next site check sees the page.

## What does a finished draft contain?

- The body, in Markdown, with the attributions the fact check wrote in.
- Under it: what to do before publishing, the claims the fact check could not trace, and "needs an editor" when the standard was not met.
- "Page title and meta description", to paste as given. Newsletter issues and messenger posts have no address of their own and get none.
- The pictures placed under their headings, each with alt text written from its concept.
- The 2 scores, quality out of 50 and quotable out of 100, live as you edit.

Every save in the editor is a new version, never an overwrite, and a version you edited is marked as yours so the agent's own writer never touches it. Older versions hang off the head of the piece on the Content tab. What the scores check is on the [writing standard page](https://www.geomarketmap.com/docs/writing-standard).

## Why do I host the pictures myself?

Because a picture served from our storage would make your page depend on us. A generated picture is kept in private storage for 7 days, long enough to download; every read of the piece renews the link for as long as we hold the file, and the piece tells you, beside every picture, to download it and host it before publishing. A screenshot is never generated; you upload it and the piece places it.

> Where we genuinely cannot finish, we say exactly what remains and never call it done.
> 
> the product's rule on turnkey work

## What is a developer's brief, and how do I hand it over?

When a site finding is not words, such as a blocked crawler, a broken structured-data block or a missing sitemap, the fix is a brief rather than a draft: what is wrong, on which pages, what to change, and how the next check will see it fixed. "Work on it" on the Health tab writes the brief; it lands on the Content tab like a draft and downloads the same way. The [developer briefs page](https://www.geomarketmap.com/features/work-orders) shows an example.

## What should the published page carry?

- The title and description as given, 15 to 65 and 50 to 160 characters.
- A visible last-updated date, and Article markup with the author and both dates on an article.
- The pictures on your own host, with the alt text kept.
- A link from at least 1 other page of yours, so the page is not an orphan.

The next site check reads the page as the kind of page it is and reports whether it is quotable; the [site check page](https://www.geomarketmap.com/docs/site-check) says what it looks for.

## Can the product publish to WordPress or another CMS for me?

No, and by design. Nothing is ever published to a customer's site on their behalf; the piece is yours to place, and the product tells you exactly what remains.

## What format is the download?

Markdown, which pastes into most editors, or a Word file for a team that works in documents. The stored format is Markdown, so the editor never becomes the format.

## Read next

- [What is a GEO MarketMap draft checked against before I see it?](https://www.geomarketmap.com/docs/writing-standard)
- [What does the GEO MarketMap site check read, and how does it judge a page?](https://www.geomarketmap.com/docs/site-check)
- [Where are the GEO MarketMap settings, and what does each one change?](https://www.geomarketmap.com/docs/settings)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/publish-a-draft


# Where are the GEO MarketMap settings, and what does each one change?

1 screen, Project settings, in 2 halves. Project: general (the market and language the material is written in), knowledge (the 5 stores the agent reads), sources, integrations (Google Search Console), and visuals (the look of a piece's pictures). Account: your account, including deletion; plans and credits, including auto top-up; and appearance, including the interface language. Every setting takes effect on the next turn or the next piece.

## What is under Project?

| Section | What it holds | What it moves |
| --- | --- | --- |
| General | The project's name and site; the market and language the material is written in, proposed from the site at the first check. | Every draft, question, title and llms.txt written after the change. |
| Knowledge | The project card, your notes, the rivals, writing rules, content templates. | The agent's next turn; see the guide on correcting what it knows. |
| Sources | Where the measurements come from, as the product reports them. | Nothing to set; it is a read. |
| Integrations | Google Search Console, connected through Google's own consent screen after the first draft exists. | Your own search data joins the demand reading. |
| Visuals | 2 looks: the cover that travels with a piece, and the pictures inside it. Each is a description of up to 300 characters and up to 3 reference images. | Every picture made after the save is drawn in that look. |

With reference images a picture is drawn from them; with a description alone the description replaces the house look; with neither the house look applies. The kind's own rules hold either way: legible labels on a diagram, no text on a cover.

## What is under Account?

- Your account: the address you sign in with, the legal texts you accepted, and "Delete my account", which says first what is lost and asks you to type your email to confirm.
- Plans and credits: the balance, the payments, and auto top-up, which is off by default. When on, a debit that takes the balance under your threshold buys the amount you chose, within the monthly cap you set; 3 failed charges in a row switch it off and the card says why.
- Appearance: the theme, and the interface language. The default follows the browser; the override is per browser, like the theme.

> Destruction asks first and says what it costs. Anything that narrows what the agent knows gets the same treatment.
> 
> the fifth of the product's 5 rules for project knowledge

## Where do schedules live?

On the assistants tab, on the Business plan: a weekly or monthly assistants check, charged like a manual run, with unanswered questions refunded. Nothing else runs on a clock. The site check and the strategy are revisited on the agent's offer after 30 days; a catalog read is offered again after 7 days. The plans are compared on the [pricing page](https://www.geomarketmap.com/pricing), and the 3 languages are explained on the [languages page](https://www.geomarketmap.com/docs/languages).

## Is auto top-up a subscription?

No. Nothing is charged on a date. A top-up happens only when a debit takes the balance under the threshold you set, within the monthly cap you set, and it is off until you switch it on.

## What happens when I delete my account?

The projects, reports, drafts and tracked questions are deleted, and the balance is lost. The screen lists this before you confirm by typing your email.

## Read next

- [How do I correct what the GEO MarketMap agent believes about my project?](https://www.geomarketmap.com/docs/knowledge-and-rivals)
- [In which language does GEO MarketMap work?](https://www.geomarketmap.com/docs/languages)
- [How does GEO MarketMap charge, and when is a charge refunded?](https://www.geomarketmap.com/docs/credits-and-billing)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/settings


---

# Reference

The terms, the billing, the languages.

# What do the terms in GEO MarketMap mean?

Each term below is defined in 1 sentence, in the sense the product uses it. The 3 that matter most: a brand is named when an answer mentions it; a gap is a question where a rival is named and you are not; and every figure is a count out of the answers collected, which is its denominator. The definitions are the same ones the agent works to.

## What are the measurement terms?

| Term | Meaning |
| --- | --- |
| Named | Your brand appears in an assistant's answer or Google's AI answer to a question. "Named in 12 of 30 answers" is the form every figure takes. |
| Cited | Your page is listed as a source the answer rests on. A page can be cited without the brand being named, and the other way round. |
| Denominator | The number of answers collected for a figure, always stated with it. Yours is your own question set, never a market. |
| Tracked question | A question the assistants are asked on every check. Agreed in step 3, changed on 1 card afterwards. |
| Niche | Every question the category asks across you and your rivals, 1 row each, judged as relevant, adjacent or off-topic. |
| Gap | A niche question where a rival is on the answer and you are not. The queue's first lane. |
| Rival | A brand the answers name instead of you, kept by you, measured 1 at a time onto the market map. |
| Market map | 1 map of the category: you, the rivals, every niche question, and who is on each answer. |
| What the answer rests on | The sites an answer cites, each with what it is: a comparison page, a review site, a forum, a rival's own page. |

## What are the planning terms?

| Term | Meaning |
| --- | --- |
| Lane | 1 of 3 kinds of work: search demand, the answers, the category's signals. Every proposal belongs to 1. |
| Strategy mix | The split of effort across the 3 lanes in percent, proposed by the agent, decided by you, revisited after 30 days. |
| Queue | 1 ordered list of everything writable, each entry with its lane, evidence and weight. |
| Move | The 1 thing the agent proposes next, by name, with its reason. 2 alternatives sit under it. |
| Catalog | A theme and the accounts that publish on it, filled by you, read on your press. |
| Probe | The free read of how alive a theme is on the platforms when a catalog is created. |
| Subject | 1 thing a catalog is talking about this week, with its kind, summary and verbatim quotes. |
| Heat Score | A number per subject from its share of posts, reach and engagement, with a velocity and a stage. Published at https://www.geomarketmap.com/heat-score. |

## What are the writing terms?

| Term | Meaning |
| --- | --- |
| Quality | A score out of 50 on whether a person would have written the piece; red under 35. |
| Quotable | A score out of 100 on whether an assistant would lift an answer from the piece; ready at 70. |
| Fact check | The pass that traces every figure and named claim to an independent source and names the ones it could not. |
| Needs an editor | The flag on a piece whose standard was not met, or whose superlative or comparison nothing stands up. |
| Version | 1 saved state of a piece. A save is a new version, never an overwrite; a version you edited is marked as yours. |
| Brief | The document handed to a developer when a fix is not words: what is wrong, where, what to change, how it will be seen fixed. |
| Cover and material | The 2 looks a project sets for its pictures: the cover that travels with a piece, and the pictures inside it. |

## What are the site terms?

| Term | Meaning |
| --- | --- |
| Site check | The free read of the site: the root files, 12 pages chosen by kind, and the home page's speed. |
| Gate | 1 of 6 conditions that stop the site being read at all, such as the assistants' crawlers being blocked. |
| Finding | 1 thing worth fixing, with the pages it is about and whose morning it belongs in. |
| Page kind | What a page is: home, product, pricing, about, article, docs, FAQ, comparison, listing, legal, contact. Each is judged as its kind. |
| llms.txt | The file at a site's root that maps its pages for a model, in the format published at https://llmstxt.org. |

> A page that cannot state the question it answers is a page nobody searched for.
> 
> the rule every page on this site is written to

## Is a credit a term too?

Yes: the unit every action is charged in, bought in advance on either plan. How charging works is on the credits and billing page; the rates are on the pricing page.

## Why does the product say "answers collected" rather than "answers"?

Because an assistant sometimes returns no answer, and an unanswered question is refunded and left out of the count. The count that is stated is the count that was measured.

## Read next

- [What does GEO MarketMap measure, and what does "named" mean?](https://www.geomarketmap.com/docs/search-and-ai-visibility)
- [How does GEO MarketMap charge, and when is a charge refunded?](https://www.geomarketmap.com/docs/credits-and-billing)
- [What is a GEO MarketMap draft checked against before I see it?](https://www.geomarketmap.com/docs/writing-standard)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/glossary


# How does GEO MarketMap charge, and when is a charge refunded?

You buy credits in advance on 1 of 2 plans, and every action is charged in credits at the moment you press it, after the control has said what it will cost. A measurement that returns nothing is refunded; an unanswered question is refunded; a job that cannot finish is refunded and the conversation says so. Nothing is charged on a date. The rates are on the pricing page and nowhere else.

## What are the rules of a charge?

- Said before the press, once. The control carries its cost; the conversation never repeats a price.
- Charged at the press, never in a lump. A 20-minute measurement is 1 charge at its start, not a meter.
- Refunded when nothing came back. The refund is plumbing, not a feature: you see the result, or the line saying the work could not finish and the credits are back.
- Never twice for the same reading. A rival the report already measured costs nothing to keep; a catalog read is answered from file for 7 days.
- The free things stay free: the site check, agreeing the questions, agreeing the niche, the probe, a re-check of the site.

> Functions work reliably and carry through to the result. The service's problems are the service's to solve, never the user's.
> 
> rule number one of the team's working rules, in the owner's words

## What is the difference between the 2 plans?

The rate per credit is the same on both. Business adds scheduled assistants checks and several projects on 1 account; Standard is 1 project on request. Both are prepaid, both have a floor you can start from and put more above, and credits on both stay valid for the period stated on the [pricing page](https://www.geomarketmap.com/pricing). There is no subscription on either.

## How does auto top-up work?

1. Save a card through the payment provider's own page; the card details never touch our servers.
2. Choose a threshold, an amount and a monthly cap.
3. When a debit takes the balance under the threshold, the amount is bought, within the cap, at most once every 5 minutes.
4. The turn that caused a top-up says so in the conversation, and the ledger shows the row.

Off by default. 3 failed charges in a row switch it off, and "Replace the card" is the way back on. It is not a subscription: nothing is charged on a date. Where the settings sit is on the [settings page](https://www.geomarketmap.com/docs/settings).

## Where do I see what was charged?

Plans and credits, under Account: the balance, every payment, and every debit and refund with the action it was for. A measurement's row says what it measured; a draft's row says which piece. The rates behind them are on the [pricing page](https://www.geomarketmap.com/pricing), which is the 1 place a price is written so that it can never go stale in 2.

## Do credits expire?

They stay valid for the period stated on the pricing page from the date of purchase, on both plans. The balance screen shows what you have and when it was bought.

## What happens if a measurement fails halfway?

The work is retried in smaller parts with a checkpoint after each. If it cannot finish, the charge is refunded and the conversation says so; you never see a half-result presented as complete.

## Can I get a refund on unused credits?

The Terms of Service set out how unused credits are handled when an account is closed or a sub-processor objection cannot be resolved. Write to the address in the Terms.

## Read next

- [Where are the GEO MarketMap settings, and what does each one change?](https://www.geomarketmap.com/docs/settings)
- [What happens in the first hour with GEO MarketMap?](https://www.geomarketmap.com/docs/first-project)
- [What does GEO MarketMap do with my data?](https://www.geomarketmap.com/docs/data-and-security)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/credits-and-billing


# In which language does GEO MarketMap work?

3 languages, each its own setting. The interface follows your browser, with an override under Appearance; English and Russian are available. The conversation follows what you type, and follows the interface setting when you press buttons. The material, meaning drafts, questions, titles and briefs, follows the project's market language, set under General and proposed from your site at the first check.

## Why 3 languages rather than 1 switch?

Because the 3 belong to 3 different people. The interface belongs to the person at the screen; the conversation belongs to the message; the material belongs to the project's market. A Russian-speaking owner of an English site reads a Russian interface, talks Russian to the agent, and gets English pages, English questions and an English llms.txt. 1 switch for all 3 would make that owner choose which of the 3 to get wrong.

> Interface and content language must never be one switch.
> 
> the finding the product's language design rests on, after a read of the localisation guides

## Where is each one set?

| Language | Whose | Where it is set | What it governs |
| --- | --- | --- | --- |
| Interface | The person at the screen | Browser default; override in Project settings, Appearance. Per browser, like the theme. | Menus, tabs, cards, the setup steps, the lines the product writes itself. |
| Conversation | The message | Nowhere: read off what you type. Pressed buttons follow the interface setting. | The agent's replies and the chips under them. |
| Material | The project's market | Project settings, General: market and language. Proposed from the site's declared language at the first check. | Documents, drafts, questions, titles, llms.txt, briefs: everything that lands on your site. |

Asking for 1 piece in another language is honoured for that piece and said so. Where the settings sit is on the [settings page](https://www.geomarketmap.com/docs/settings); what the writer does with the market language is on the [writing standard page](https://www.geomarketmap.com/docs/writing-standard).

## Does the measurement depend on the language?

Yes. The questions are asked in the market's language and the answers are read in it, so a project whose market is Germany is measured on German questions. The semantics model reads meaning rather than spelling, so phrases in the market's language are grouped the same way as English ones.

## Which interface languages exist?

English and Russian today. Every string in the interface is 1 row with both languages, and a missing translation fails the build, so a language that is offered is complete.

## Can 1 project write in 2 languages?

A project has 1 market language. A piece in another language can be asked for and is written in it; a second market is a second project.

## Read next

- [Where are the GEO MarketMap settings, and what does each one change?](https://www.geomarketmap.com/docs/settings)
- [What is a GEO MarketMap draft checked against before I see it?](https://www.geomarketmap.com/docs/writing-standard)
- [How does the GEO MarketMap agent decide what to do?](https://www.geomarketmap.com/docs/the-agent)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/languages


---

# Trust

Your data, our promises, and how machines read this site.

# What does GEO MarketMap do with my data?

It reads your site's public pages and public search results and assistant answers about your category; it stores what it measured and what it wrote, for as long as your account is active; only you and the people you share a link with can see a report; and nothing is ever published to your site or your accounts on your behalf. The legal texts are the Privacy Policy, the Terms and the Data Processing Agreement; this page is the plain version.

## What is read, and from where?

- Your site: its public pages, root files and speed, read the way a crawler reads them. No login, no upload.
- Public search results and Google's AI answers for your category's queries.
- Public answers from ChatGPT, Perplexity and Gemini to the questions you track.
- Public posts from the social accounts you list in a catalog.
- Your Google Search Console data, only if you connect it, only for the features you asked for, under Google's [API Services User Data Policy](https://developers.google.com/terms/api-services-user-data-policy) and its Limited Use requirements.

Nothing is read from your private systems. Everything the measurements rest on is something anybody could read.

## What is stored, for how long, and who can see it?

The measurements, the reports, the drafts and their versions, the tracked questions, the rivals, your notes and the record of every decision, while your account is active. When you close the account they are deleted without undue delay, within the period the law requires; the [Privacy Policy](https://www.geomarketmap.com/privacy) states the retention periods. A generated picture is kept for 7 days in private storage and then swept. A report is visible to you and to anybody you share its link with; shared links are kept out of search engines.

Access is per account: a project belongs to 1 account, and the product's own staff see a project's content only to support you, under the Privacy Policy. Card details are handled by the payment provider and never touch our servers.

## What does AI-generated content mean for me?

The reports, the recommendations and the drafts are produced with AI models. They are checked, scored and fact-checked before you see them, and they can still contain errors, because the platforms they measure change. You review a piece before you publish it and you decide what to act on; the product's own standard requires a person to approve everything it publishes on its own site. AI-generated content is not professional advice; the [Terms of Service](https://www.geomarketmap.com/terms) say so in full.

> Human Oversight: users must review outputs before acting upon them.
> 
> section 9 of the Privacy Policy

## Which promises hold on every screen?

- Nothing is published to your site, your social accounts or anywhere else on your behalf.
- No ranking or position is promised.
- No figure is given without the count it was measured on.
- Every cost is said before the press, and a charge for nothing is refunded.
- The suppliers and models behind a measurement are not named in the product; the sub-processors that handle personal data are listed in Annex 1 of the Data Processing Agreement, with 30 days' notice before a change.
- Deleting a rival, a question or the account says first what is lost.

Security measures, sub-processors and international transfers are set out in the [Data Processing Agreement](https://www.geomarketmap.com/dpa); a detailed security schedule can be provided under NDA. The address for data requests is in the Privacy Policy.

## Is my site's content used to train models?

No. Your pages are read to measure and to write for you, and the product does not use customer content to train models of its own. The Data Processing Agreement sets out the processing purposes.

## Can I export my data?

Every draft downloads as Markdown or Word; every report has a shareable link and the demo shape is public. Requests for a full export go to the address in the Privacy Policy.

## Where is the company established?

The Terms of Service and the Data Processing Agreement name the entity, its establishment and the governing law. The sub-processors and their locations are in Annex 1 of the DPA.

## Read next

- [How does GEO MarketMap charge, and when is a charge refunded?](https://www.geomarketmap.com/docs/credits-and-billing)
- [How does the GEO MarketMap agent decide what to do?](https://www.geomarketmap.com/docs/the-agent)
- [How should an AI assistant or crawler read geomarketmap.com?](https://www.geomarketmap.com/docs/for-ai-assistants)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/data-and-security


# How should an AI assistant or crawler read geomarketmap.com?

Start at /llms.txt, which maps the site in the format published at llmstxt.org. Read /llms-full.txt for the whole documentation as 1 Markdown file. Every documentation page has a Markdown twin at the same address with .md appended, linked from the page's head. The crawlers behind assistant answers are allowed by name in robots.txt, every page carries structured data, and every page opens with the answer to its own question.

## What files are published for models?

| File | What it is | How often it changes |
| --- | --- | --- |
| https://www.geomarketmap.com/llms.txt | The map: 1 line per page with a note, in the format at https://llmstxt.org. Small enough to sit in a context window. | Within the hour of a page shipping. |
| https://www.geomarketmap.com/llms-full.txt | The whole documentation, every page as Markdown, in reading order. | With every documentation edit. |
| https://www.geomarketmap.com/docs/<slug>.md | 1 documentation page as Markdown: the same words the HTML page renders, nothing else. | With the page. |
| https://www.geomarketmap.com/sitemap.xml | Every public address with its last-modified date. | Within the hour. |
| https://www.geomarketmap.com/robots.txt | Who may crawl. GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended are allowed by name. | Rarely. |

The files are derived from the same data the pages render, so a page that ships appears in all of them without anybody remembering to add it. The site check the product sells reports a site that lacks these; ours would fail its own check otherwise.

## What structured data does each page carry?

Every documentation page carries an Article record with its headline, publisher, publication and modification dates, and a FAQPage record with the page's question and its answer first, then the questions also asked. Breadcrumbs name the section. The vocabulary is [schema.org](https://schema.org/FAQPage), and the records are built from the same fields the page renders, so they cannot describe a page as it was 6 months ago.

> Every page opens with the answer to its own question, in the shape an assistant can lift whole.
> 
> the rule this site's pages are written to

## What may a model quote, and how should it attribute?

The documentation and the feature pages describe the product and may be quoted with attribution to GEO MarketMap and a link to the page. The Index and the research pages carry original measurements; each is published with the questions asked and the number of answers collected, and a quoted figure should carry that denominator with it. Nothing on the site names the product's suppliers or models, and a model should not infer them.

## What is deliberately not indexable?

Shared report links, the signed-in workspace, the screenshot studio and the parked previous home page are disallowed in robots.txt and carry no index directive. A shared report belongs to the customer who shared it; a model that reaches one should treat it as private.

## Do assistants actually read llms.txt?

As of 2026 no major assistant has confirmed that it uses the file to decide what to cite, and Google has said it does not use it for search. It costs nothing to publish, it is the published convention, and the Markdown twins are useful to any reader that fetches them.

## Is there an API?

Not a public one. The machine-readable surface is the files above, and the product itself is used through its interface and its agent.

## Read next

- [What does the GEO MarketMap site check read, and how does it judge a page?](https://www.geomarketmap.com/docs/site-check)
- [What does GEO MarketMap do with my data?](https://www.geomarketmap.com/docs/data-and-security)
- [What do the terms in GEO MarketMap mean?](https://www.geomarketmap.com/docs/glossary)

Last updated 15 September 2026. Source: https://www.geomarketmap.com/docs/for-ai-assistants