Strategy

Short answer: B2B companies get cited by ChatGPT, Perplexity and Gemini when their pages are accessible to AI search crawlers, answer specific questions clearly, contain citable facts such as original data and definitions, and when credible third-party sources describe them consistently. Being mentioned on the review sites, directories and publications that AI tools draw on matters as much as your own site.

How B2B companies get cited by ChatGPT, Perplexity and Gemini, cover

Buyers increasingly ask AI tools for recommendations and explanations before they search or contact a vendor: which CRM suits a small manufacturer, what a fractional CMO costs, how to set up server-side tracking. When those answers mention or cite your company, you reach buyers at an early, influential moment. Nobody outside the AI companies knows exactly how each tool chooses sources, and the methods change. But the tools' own documentation and observable behaviour point to a consistent set of factors. This article explains how AI answer tools generally choose sources, the work on and off your site that makes citations more likely, and how to track progress.

How AI answer tools find sources

AI assistants answer from two kinds of knowledge. The first is what the model learned during training, which reflects the web and other text up to a cut-off date. The second is what the tool retrieves at the time of the question, by searching the web or an index and reading pages before writing the answer. ChatGPT search, Perplexity, Gemini and Google's AI features all use retrieval for many questions, and when they do, they often show citations linking to the pages they used. To be cited, your pages need to be retrievable by these systems and to contain information that answers the question better or more specifically than alternatives. To be mentioned without a citation, your brand needs to be present in the sources the model learned from or retrieves.

Let the right crawlers in

Check that AI search crawlers can access your site. OpenAI documents separate user agents for search (OAI-SearchBot) and for training (GPTBot), so you can allow search access while making your own decision about training. Perplexity documents PerplexityBot. Google's AI features use Google's existing search crawling, with a separate control, Google-Extended, for some uses of content in Gemini models. Review robots.txt, firewall and bot-protection settings, because security tools sometimes block these crawlers by default. The articles on GPTBot vs OAI-SearchBot and on tracking AI crawlers in server logs on this site explain the tokens and how to confirm crawlers are reaching you.

Make pages easy to cite

AI tools tend to cite pages that answer a specific question directly and contain something worth quoting. Put a clear answer of a few sentences near the top. Use headings that match the questions buyers ask. Include specific, checkable information: definitions, steps, prices or price ranges where you can publish them, comparison tables, original data, worked examples and named frameworks. Cite your own sources for facts. Keep pages up to date and show the date. Generic pages that say what every competitor says give a model little reason to choose yours over a better-known source.

Off-site sources matter as much as your site

AI answers frequently draw on third-party pages. Check where your buyers' questions are answered and whether you appear accurately there.

Source types AI tools often draw on, and what to do about each
Source typeExamplesWhat to do
Review and comparison sitesSoftware review platforms, agency directoriesComplete profiles, gather genuine reviews, keep details current
Industry publicationsTrade media, newsletters, association sitesContribute expert articles, data and commentary
Community discussionsForums, Q&A sites, professional communitiesParticipate helpfully under your real name; never fake posts
Lists and roundupsBest tools for X, top consultants in YIdentify the lists AI cites for your queries; earn inclusion with merit and outreach
Reference and profile pagesBusiness listings, professional profiles, partner directoriesUse one consistent name, description and service list
Video and podcastsYouTube, podcast transcriptsPublish talks and interviews with clear titles and descriptions

Be one consistent entity

AI tools need to connect mentions of your company across the web. Use the same name, description, location, services and key people everywhere: your website, structured data, social profiles, directories, partner listings and author bios. Link them with sameAs properties in your organisation structured data. Inconsistent names, such as a brand name on one site and a legal name on another, or outdated service descriptions, make it harder for a model to attribute information to you. The entity SEO article on this site sets out a full checklist.

Publish things only you can publish

Original data, benchmarks, calculators, case studies with real numbers and clear positions on contested questions are more likely to be cited than summaries of other people's work, because they are primary sources. A short benchmark report based on your own anonymised data, a public calculator or a well-sourced comparison can become the page that AI tools quote for a topic. Make such assets easy to reference: a clear title, a date, a methodology note and stable URLs.

Checking progress

Build a set of prompts that represent buyer questions, such as best options, comparisons, costs and how-to questions in your category, and run them in ChatGPT, Perplexity, Gemini and Google regularly, ideally monthly. Record whether your brand is mentioned, whether your pages are cited, what is said about you and which sources are cited instead. Track AI referral traffic in analytics. Use the cited sources as a to-do list: if a particular directory or article is cited repeatedly for your category, make sure you are accurately represented there. The article on tracking your brand in AI answers covers how to build the prompt set.

What not to do

Avoid tactics that try to manipulate models: hidden text aimed at AI crawlers, mass-produced pages targeting every possible prompt, fake reviews or forum posts, and paid placements disguised as editorial content. These conflict with search and platform policies, can harm your reputation if discovered, and tend not to last as tools improve. Durable visibility comes from being genuinely useful, accurately described and referenced by sources buyers trust.

A 90-day plan

A focused quarter can make real progress. In the first month, check crawler access, fix robots.txt and firewall settings where needed, update organisation and person structured data, and build the prompt set to record a baseline. In the second month, restructure your ten most important pages answer-first, with question headings, a comparison table or worked example and sources, and update or complete your profiles on the review sites and directories that appear in the baseline answers. In the third month, publish one primary asset, such as a benchmark, calculator or detailed guide built on your own experience or data, and pitch it to the publications and newsletters your buyers read. At the end of the quarter, rerun the prompt set and compare with the baseline. Expect gradual change rather than a sudden jump, and keep the routine going, because AI answers shift as models and sources change.

Measuring the commercial effect

Citations and mentions matter only if they reach buyers who act. Track referral traffic from AI tools in analytics, and look at what those visitors do: pages viewed, enquiries and conversions. Ask new leads how they researched options, with AI assistants as one of the choices, and record the answer in the CRM. Compare the quality and conversion of leads who mention AI tools with other sources. Over time, this shows whether AI visibility is producing pipeline and which topics and pages drive it, so effort can focus on the questions that bring buyers rather than those that bring only curiosity.

FAQ

Allow OpenAI's search crawler, publish pages that answer specific questions with citable facts, keep your entity information consistent and be accurately represented on the third-party sites AI tools draw on.

OpenAI documents separate user agents for search (OAI-SearchBot) and training (GPTBot). Blocking GPTBot does not by itself block OAI-SearchBot. Check OpenAI's current documentation.

Often. AI tools frequently draw on review sites, directories and roundups, so accurate, complete profiles and genuine reviews help.

It varies by tool and topic. Retrieval-based answers can change as soon as new pages are indexed; mentions based on training data change more slowly. Track monthly with a fixed prompt set.

Paid ad formats may exist in some tools, but organic citations cannot be bought. Treat any offer of guaranteed organic AI citations with caution.

Track AI referral traffic and its conversions in analytics, ask new leads how they researched options and compare lead quality from AI-assisted research with other sources.

Pages that answer a specific question directly and contain citable specifics, such as definitions, comparisons, prices, original data and worked examples.

Read this article on your favourite platform

Ready to build the system?

If this describes your funnel, a 30-minute call will find where the constraint sits in your own numbers and what it would take to fix it.

It starts with a 30-minute call. Pick a time below.

Prefer email?