The AI discovery checklist.

What is AI discovery, and why does it matter now?

AI discovery is how buyers find and shortlist a company through an AI assistant instead of a page of search results. To be discoverable, a company has to be easy for the assistant to identify, easy to describe accurately and backed by pages and third-party sources it can cite. Generative engine optimisation (GEO) is the name for that work; the thinking behind it is in our article on generative engine optimisation and AI search. This guide is the practical companion: a checklist in five groups, in the order we would work through them.

The shift is measurable. In a survey of more than 1,000 B2B software buyers in August 2025, G2 found that half now start their research in an AI chatbot rather than Google. Pew Research Center’s 2025 study of US browsing data found that users clicked a traditional result link in 8% of visits when a Google AI summary appeared, against 15% when none did. Fewer clicks means the answer itself is where the shortlist forms.

1. Is your company a clear, consistent entity?

An assistant can only recommend a company it can identify without ambiguity. Start by making the name, the one-line description and the core facts identical everywhere they appear.

• One name, one description. Use the same legal and trading name, the same one-sentence description of what the product does and for whom, and the same category words on the website, social profiles, review sites and press boilerplate.
• Organization structured data. Add schema.org Organization markup to the home page with name, logo, URL, founding date and sameAs links to official profiles. Google documents the fields it reads in its Organization structured data guide.
• A factual About page. What the company does, who it serves, where it operates, who leads it and when it was founded, written as plain statements an assistant can lift.
• Wikipedia and Wikidata, only where notable. Wikipedia’s notability guideline for companies requires significant coverage in independent sources, and its conflict-of-interest rules discourage writing about your own company. If the coverage exists, make sure existing entries are accurate; if it does not, build the coverage first.
• Complete profiles. G2, Capterra, LinkedIn and Crunchbase profiles with the same description, category, headquarters and website as everywhere else.

2. Does your content answer the questions buyers ask?

Assistants quote pages that answer a question directly. Write each important page around one question and put the answer in the first two or three sentences.

• Question-led pages. Build pages around the questions buyers type into assistants: “What is ...?”, “How does ... work?”, “Which ... is right for a team of 50?”. Use the question, or close to it, as the heading.
• Direct answers first. Open each section with a one or two sentence answer, then the detail, evidence and caveats. Define every term plainly the first time it appears.
• Comparison and alternatives pages. Honest, factual pages comparing your product with named alternatives, including where the alternative is the better fit. Buyers ask assistants for comparisons; give them a source that is fair.
• Pricing pages. Publish how pricing works, even when the final figure needs a conversation. A page that says nothing gives the assistant nothing to quote, and it will quote someone else.
• FAQs with FAQPage markup. Self-contained answers with FAQPage structured data. Note that Google said in August 2023 that FAQ rich results would only be shown regularly for well-known, authoritative government and health websites, so treat the markup as a clear description of the page, not a route to rich results.
• A glossary. Short definitions of the terms in your category, each on a stable URL, written so a sentence can be quoted on its own.

3. Can AI crawlers reach and read your site?

An assistant cannot cite a page its crawler cannot fetch. Check robots.txt, rendering and sitemaps before anything else technical, and know which crawler controls what.

• OAI-SearchBot and GPTBot are separate. OpenAI’s crawler documentation says OAI-SearchBot surfaces websites in ChatGPT search, and sites opted out of it will not be shown in ChatGPT search answers. GPTBot crawls content that may be used to train OpenAI’s foundation models. ChatGPT-User makes visits a user triggers, and robots.txt rules may not apply to it.
• PerplexityBot and Perplexity-User. Perplexity’s documentation says PerplexityBot surfaces and links websites in its search results and is not used to crawl content for AI foundation models. Perplexity-User fetches pages when a user asks a question and generally ignores robots.txt.
• Google-Extended is not a Search switch. Google’s crawler documentation describes Google-Extended as a token to manage whether crawled content may be used for training future Gemini models and for grounding, and says it does not affect inclusion or ranking in Google Search. AI Overviews and AI Mode draw on the normal Search index; Google’s guidance on AI features says there are no additional requirements beyond being indexed and eligible for a snippet.
• Decide training and search separately. A common pattern is to allow the search crawlers and make a deliberate, documented choice about the training crawlers. Check the CDN or firewall too: bot protection can block crawlers that robots.txt allows.
• Server-rendered HTML. Make sure the main content, headings and links are in the HTML the server returns. Do not assume every crawler will run your JavaScript the way a browser does.
• Sitemaps and speed. An XML sitemap with accurate last-modified dates, referenced in robots.txt, and pages that load quickly and return clean status codes.
• llms.txt, honestly. llms.txt is a proposal published by Jeremy Howard in September 2024: a Markdown file at the site root with a summary and links for AI agents. It is an emerging convention, not a standard, and Google says no special machine-readable AI files are needed for its AI features. It is cheap to add, and worth doing only after everything above is in place.

4. What do other sources say about you?

Assistants draw on the wider web, not only on your own site. Independent evidence is what lets an assistant describe a company with confidence and name it in a shortlist.

• Reviews. A steady flow of recent, specific reviews on the review sites your buyers use, with replies from the team. Ask customers at natural moments, such as after onboarding or a renewal.
• Analyst and press mentions. Briefings with the analysts who cover your category, and coverage in trade press that states plainly what the company does.
• Partner pages. Listings in integration marketplaces and on partners’ websites, with the same description you use elsewhere.
• Community answers. Useful, disclosed answers from your team in the forums and communities where buyers ask for recommendations. Disclose the affiliation every time; undisclosed promotion damages trust and usually breaks the community’s rules.

5. How do you measure visibility in AI answers?

Measure three things: referral traffic from assistants, how often you appear in answers to the questions that matter, and whether those answers are accurate.

• AI referrals in analytics. Group referrals from assistant domains such as chatgpt.com, perplexity.ai and gemini.google.com into their own channel. OpenAI’s publishers and developers FAQ says ChatGPT adds utm_source=chatgpt.com to referral URLs. Google counts AI Overviews and AI Mode traffic inside overall Search traffic in Search Console, so it cannot be isolated there.
• Prompt testing. Keep a fixed list of 20 to 50 buyer questions, from category definitions to “alternatives to” and pricing questions. Run them in each major assistant every month and record whether you appear, how you are described and which sources are cited.
• Share of AI answers. From that log, track the proportion of answers that mention you against named competitors. It is a directional measure: answers vary between runs, so look at trends over months, not single results.
• Citation monitoring tools. A growing set of commercial tools automates prompt runs and tracks citations at scale. Before paying for one, check which assistants and regions it covers and how it samples answers.
• Accuracy checks. When an assistant gets a fact wrong, trace it to the source it cites and correct that source, whether it is your own page, a profile you can update or a third party you can contact.

Where should a founder start?

Start with crawl access and entity clarity, because every other step depends on them, and checking robots.txt, firewall rules and profile consistency is quick work. Then rewrite the ten pages buyers most need answered, set a prompt-testing baseline, and build third-party evidence quarter by quarter. Treat the work as part of go-to-market, owned by the same people who own pipeline, not as a side project for the website.

TGC Capital Partners, Your Growth Engineering Partner, works on this as part of 1X Revenue Engineering, where discoverability sits alongside positioning, pipeline and sales execution, and with 2X Software Engineering & Tech Excellence when the fixes are technical. Revenue is the outcome we measure. The wider model is on the 6X Factors platform page.

Frequently asked questions

What is AI discovery?

AI discovery is the process by which a buyer finds and shortlists a company through an AI assistant such as ChatGPT, Perplexity, Gemini or Google AI Overviews, rather than through a list of search results. Being discoverable means the assistant can identify the company correctly, describe what it does accurately and cite its pages or third-party evidence about it.

Does blocking GPTBot remove a site from ChatGPT search?

No. OpenAI documents GPTBot as the crawler for training its foundation models and OAI-SearchBot as the crawler that surfaces websites in ChatGPT search. A site can disallow GPTBot for training and still allow OAI-SearchBot so that its pages can appear in ChatGPT search answers.

Does a software company need an llms.txt file?

It is optional. llms.txt is a proposal published by Jeremy Howard in September 2024, not a ratified standard, and Google states that no special machine-readable AI files are needed to appear in AI Overviews or AI Mode. It is cheap to add, but it should come after crawl access, clear pages and structured data, not instead of them.

Is FAQPage structured data still worth adding?

Yes, as a description of the page rather than as a route to rich results. Google said in August 2023 that FAQ rich results would only be shown regularly for well-known, authoritative government and health websites. The markup still states plainly which questions a page answers, which helps any system that reads structured data.

How can a company see traffic that comes from AI assistants?

Create a channel or segment in web analytics for referrals from assistant domains such as chatgpt.com and perplexity.ai, and watch for the utm_source=chatgpt.com parameter that OpenAI says ChatGPT adds to referral links. Google counts visits from AI Overviews and AI Mode inside overall Search traffic in Search Console, so those are not reported separately.

How long does AI discovery work take to show results?

There is no fixed timeline and no outcome can be promised. Crawler changes can take a day or more to be picked up, Google notes that recrawling can take several days to several months, and third-party evidence such as reviews and press builds over quarters. Founders should set a baseline of prompt tests first and re-run them on a fixed monthly cadence.

Speak with a partner 1X Revenue Engineering

Related reading: Generative engine optimisation and AI search · What turns SaaS innovation into recurring revenue? · Rebranding for growth · Geographic expansion playbook