GEO consulting: get your brand cited in AI answers.

GEO (generative engine optimization) is the work of getting your brand cited as a source in answers produced by artificial intelligence (AI), and by the more capable systems now discussed as superintelligence (ASI): Google’s AI Overviews and AI Mode, ChatGPT search, and answer engines such as Perplexity and Bing Copilot. We wrote this guide for marketing teams who know SEO but can’t work out why they don’t show up in AI answers. Every section tells you what to change on your site and in your brand’s footprint across the web.

In GEO consulting, our job is to find the questions you should be cited for, shape your site and content so AI can read and quote it, grow your brand’s footprint on the web, and measure citations on a regular schedule. The sections below explain how each of these four jobs is done. The same work is also sold as “AI SEO” or “AI search optimization”; whatever the label, what matters is who sets up the measurement and who makes the changes. For a short definition and where the term comes from, see our post What is GEO?

GEO vs SEO: is AI SEO a separate discipline?

SEO aims to rank your site in search results; GEO aims to get you cited as a source inside the answer an AI engine gives. Both sit on the same foundation: a crawlable site, content that answers questions directly, and a trustworthy brand. The difference comes from the output you’re after: not a position in the rankings, but a mention in the answer.

Don’t read this table as two separate disciplines. Everything you do for GEO helps SEO too. According to Google, for a page to appear as a supporting link in AI Overviews or AI Mode, it must be indexed and eligible to be shown in Search with a snippet;¹ a page that isn’t indexed can’t be a source in an AI answer either.

Where do AI answers appear?

Four surfaces stand out today. On some searches, Google places an AI Overview above the results, and it also offers a conversational AI Mode. ChatGPT searches the web for current questions and adds source links inside its answer. Perplexity gives every answer with numbered sources. Bing Copilot draws on the Bing index. Google states that there are no additional requirements or special optimizations for appearing in these features beyond the fundamentals of SEO.¹ Bing likewise recommends clear, well-structured content that puts the key information at the top of the page.²

What these surfaces share is this: the answer is built from a passage taken from your page, or from a link to it. To appear in the answer, your page has to be found first, then read, then chosen as the clearest source for that question. GEO tackles these three steps in order.

How does source selection work?

Answer engines draw on two sources: the model’s training data and a web search run at query time. For current and commercial questions, the second one decides. The engine runs a few searches close to the question, reads the pages that come back, builds the answer from them and shows the passages it quoted as sources. That’s why pages that are strong in the search index are also the closest to becoming sources in AI answers.

Structured data and a clean heading hierarchy help machines understand a page, but they don’t create visibility on their own: Google says no special markup is needed for its AI features,¹ and Bing writes that structured data doesn’t guarantee visibility.² Your brand’s footprint is the other half of the work: when an answer engine searches the web and sees your brand described accurately on news sites, in industry directories and in comparison articles, it’s easier for the answer to describe you correctly.

How do you build content that answers?

On the content side, the job is to find the questions your audience asks AI and build the page that gives each one the clearest answer. The order below is the structure we use for a service or product page.

1. Question map: we collect the questions customers ask from your sales team, support tickets and search data; comparison, pricing and “which is better” questions go first.
2. Heading = question: each section’s heading is the question itself, or very close to it; readers and engines alike can tell from the heading what the page answers.
3. First sentence = answer: the first sentence under the heading gives the answer; conditions and exceptions come in the sentences after it (Bing’s “key information first” advice).²
4. Tables and lists: answers that need a comparison or steps are given as a table or a numbered list; the parts of the answer become easier to tell apart for people and machines.
5. Date and source: when the page was updated and what each claim rests on are stated clearly; readers and engines can both see whether the information is current.
6. One topic, one page: a page answers one question in depth; splitting the same topic across several similar pages makes it unclear which one should be taken as the source.

Blog posts built on this structure work both in search results and in AI answers. Reworking existing content to the same pattern usually pays off faster than writing from scratch.

A machine-readable site: structured data, llms.txt, robots

There are three jobs on the technical side. The first is structured data: marking up organization, service, product and article information with the schema.org vocabulary,⁴ matching exactly what’s visible on the page.⁵ The second is the rules given to crawlers: robots.txt says which bot may read what.⁶ llms.txt, meanwhile, is a short map a site offers to language models; it’s an independent proposed standard,⁷ and Google writes that no such file is needed for its AI features.¹ Because it costs little, we set it up with other tools in mind; markention’s AI visibility analysis generates a draft. The third is crawl health: pages that load fast, have no redirect chains and resolve to a single canonical URL.

Which bots to allow is a brand decision. If you want to be cited in answers, you need to be open to search bots (Googlebot, Bingbot, OAI-SearchBot, PerplexityBot, Claude-SearchBot); permissions to supply data for model training (GPTBot, ClaudeBot, Google-Extended) are weighed separately. Because visitors arriving on a user’s request don’t always follow robots.txt,⁹¹⁰ we protect sensitive content with access control, not robots.txt. We make this decision with you and write robots.txt and llms.txt to match. How we audit and set these rules bot by bot is covered on our AI access control page.

Brand footprint: mentions and authority

When an answer engine searches the web, it sees your brand not only on your own site but also on news sites, in industry directories, in comparison articles and in listings. Being there with accurate, current information helps the answer describe you correctly. That’s why one leg of GEO work is digital PR: taking your brand, product and expertise to genuine third-party sources. We don’t resort to artificially manufactured mentions or links.

In practice, we find the directory and comparison articles in your industry and fill in the places where your brand is missing; we make sure company details (name, address, domain, social accounts) are identical everywhere; we set up author pages and expertise signals; and we make customer reviews and case write-ups easy to reach.

How do you measure citations?

Measurement in GEO is different from rank tracking; two people can get different answers to the same question. So we measure with a question set: we define the questions your audience asks, run the same set on different assistants at regular intervals, and record whether your brand and your competitors are cited, which page is quoted and in what context. The effect of content changes is read from these records.

The report looks at three things: citation rate (how much of the question set you appear in), source share (how many of the quoted pages are yours) and context (a recommendation, a warning, or just a name). On the site side, AI crawler access, structured data errors and crawl logs are monitored; the AI performance report in Bing Webmaster Tools (preview) adds data for the Copilot side.³

What should a GEO specialist know?

A GEO specialist is a good SEO specialist first; visibility in AI answers is built on pages that are indexed, crawlable and answer the question clearly.¹ On top of that, they need to know how answer engines choose sources, the difference between being cited and merely being named, and how to measure both. The table below sets the knowledge we look for in a GEO specialist next to the mistake we see most often in the field.

Tools used in GEO work

GEO measurement happens inside the answer engines themselves; diagnosis on the site side is completed with Search Console, Bing Webmaster Tools and analytics. No single tool shows every answer engine, so the same question set is run across several engines.

The top strip of the screen summarises one analysis run: mention rate (in how many answers across the question set the brand is named), source citations (answers that link to the site), share of voice (our share of all brand mentions) and average position in the answers where we appear. Each value is compared with the previous analysis. Below sit share of voice by brand, the breakdown by engine and change over time; further down the screen, concrete suggestions are listed per question in the form “a competitor is cited, we aren’t”, and a suggestion becomes a task in one step. The team’s AI assistant, Limon, summarises the measurements and suggests priorities; the specialist decides which page changes.

Measuring mentions is not enough on its own; we also track whether answers bring visits to the site. This card splits GA4 referrals by AI assistant and places the Google AI Overview status from keyword tracking next to it: the gap between the number of keywords with an overview and the number where we are a source shows where the next piece of work is.

Common mistakes

The GEO consulting process: the first 30 days

1. Week 1 · Measurement: the question set is built and the first citation report is taken on all four surfaces; AI crawler access, structured data and llms.txt are audited.
2. Week 2 · Plan: for the questions you aren’t cited for, we decide which pages to build or rework and which third-party sources to approach.
3. Week 3 · Implementation: robots.txt and llms.txt are written, structured data is set up and the first pages are turned into an answer-first structure.
4. Week 4 · First reading: the question set is run again; changed citations, quoted pages and questions still uncovered are reported, and next month’s plan comes out of it.

In the months after that, the work continues with content production and brand footprint work; the citation report is refreshed every month.

GEO and AI SEO pricing: what drives the cost?

The price of GEO consulting depends on the size of the question set, how many answer engines and languages are measured, how many pages are reworked, whether content production and brand footprint work are in scope, and the monthly effort assigned to the project. Instead of a fixed price list, we share estimated ranges; we work out the scope together in the first meeting and send a written proposal.

• Question set and engines: measuring 20 questions on four engines is not the same job as measuring 100 questions in three languages.
• Measurement frequency: every analysis run means paid calls to the answer engines; the number of runs drives both effort and direct measurement cost.
• Technical setup: the current state of robots.txt, llms.txt, structured data and crawl health.
• Content and brand footprint: restructuring existing pages to answer questions, new content, and visibility work on third-party sources.
• Competition: how established competitors already are in the answers.

For GEO consulting we work on a monthly effort commitment; a one-off audit and setup can be discussed separately from monthly monitoring. The table below shows estimated ranges; the final fee is set in a written proposal once the question set and scope are clear.

GEO consulting pricing: estimated ranges

Estimated range, excluding VAT; rows marked “/ month” are monthly, the others are one-off. Scope is confirmed in a written proposal after the first meeting. As of October 2026.

AI SEO agency, independent consultant or in-house team?

All three models answer the same two questions differently: who measures the answer engines, and who changes the pages. Many firms calling themselves an “AI agency” build chatbots or automation rather than search visibility, so check whether a proposal actually includes citation measurement.

We combine consulting with delivery: we build the question set and the measurement, then handle the technical fixes and the content; if your subject-matter expert sits in your team, we gather the knowledge from them.

How to choose a GEO agency

GEO is a new field; you need a partner who moves forward by measuring and testing rather than promising hard rules. When choosing between an independent consultant, an agency and a consulting firm, we suggest looking for concrete answers to these questions: how are mentions measured, which answer engines are covered, who does the content and technical work, what will be done for the brand footprint, and what will you see at the end of the first month? “We optimize for superintelligence” isn’t an answer.

1. Measurement method: the question set is written down and built from unbranded questions; every run repeats the same questions and is compared with the previous one.
2. More than one engine: no single assistant is enough; Google AI Overviews, ChatGPT, Gemini, Perplexity and Copilot are tracked separately.
3. Guarantee language: “we’ll get you to the top in ChatGPT” or a promised mention rate is an unmeasurable promise; no one can guarantee from the outside how answer engines choose their sources.
4. Access and ownership: Search Console, Bing Webmaster Tools, GA4 and the site stay in your accounts; the partner is added as an authorised user.
5. Bot decisions: robots.txt changes are made with your approval and bot by bot; training bots and search bots are assessed separately.⁶
6. Brand footprint methods: the third-party work is explained; manufacturing mentions, fake reviews or forum posts doesn’t work and harms the brand.
7. What the report contains: mentions, source share and AI-referred visits, alongside the work done that month.
8. Exit terms: the question set, the measurement history and the list of work done are handed over when the engagement ends.

You are welcome to use this list when you talk to us as well: we write the question set with you, open the accounts in your name and don’t guarantee mentions; every month we report the same question set against the previous measurement. Since GEO stands on the same foundation as search, our SEO consulting page explains that side of the work.

To see where your site stands today, you can run our free AI visibility analysis: it reports on page types, structured data, crawler access and llms.txt, and arrives in your inbox. If you’d like to talk about your own situation, pick a meeting day below.