On this page
Picture your next buyer. Before they visit your website, before they fill in a form, they open ChatGPT and type: “Which agencies can help a mid-size SaaS company fix its HubSpot lead routing?” The assistant gives them four names. If yours isn’t one of them, you were never in the running, and your analytics will never show it.
This is no longer a thought experiment. According to Forrester’s 2026 B2B buyer journey research, nearly three-quarters of B2B software buyers consult ChatGPT at some point during vendor evaluation, and 44% use Perplexity while building a shortlist. Yet Crackle PR’s Q2 2026 AI Citation Benchmark found that 51% of B2B tech brands get no citations at all across ChatGPT, Perplexity and Gemini.
The good news: most of what makes a company visible to AI assistants is within your control, and much of it is the same work that makes a website good for human buyers.
How AI assistants decide who to mention
AI assistants draw on two sources when they answer a question about vendors.
- What the model already knows. Large language models learn from a huge snapshot of the public web. If your company was described clearly and consistently across many sites, the model has a picture of what you do. If it wasn’t, the model may know nothing, or something out of date.
- What it looks up right now. ChatGPT’s search mode, Perplexity and Gemini fetch live web pages to answer many questions, then cite them. Here, the assistant behaves a lot like a search engine combined with a very fast reader: it finds candidate pages, reads them, and picks the ones that answer the question best.
You influence the first slowly, through your reputation across the web. You influence the second directly, through your own website.
What makes a page easy for AI to recommend
Answer specific questions on specific pages
A homepage that says “we deliver innovative digital solutions” gives an assistant nothing to quote. A page titled “HubSpot lead routing for B2B SaaS” that explains the problem, the approach, typical deliverables and who it’s for gives it everything.
The single most effective change is usually structural: one page per service and per buyer question, each opening with a plain, direct answer in the first paragraph.
Say who you are, in the same words everywhere
Assistants piece together a picture of your company from many sources. Help them by describing yourself the same way on your website, LinkedIn, directories and partner listings: what you do, for whom, and where. Inconsistent descriptions (“web agency” here, “AI consultancy” there) produce a blurry picture, and blurry companies don’t get recommended.
Be specific and verifiable
AI systems favour content that contains concrete facts: named tools, clear processes, real numbers with sources, and dates. Vague claims are hard to quote and easy to ignore. Unsourced claims, like an unexplained “5-star rating”, can also undermine trust with the human reader who checks.
Get mentioned on other people’s sites
Assistants trust what others say about you more than what you say about yourself. Reviews on directories like Clutch or G2, listings in partner directories (HubSpot, Clay, n8n), articles you contributed to, podcasts and conference talks all create independent mentions. For many B2B firms this is the biggest gap.
Use structured data and clean technical basics
- Structured data (schema.org markup) for your organisation, services, articles and FAQs helps machines understand what each page is.
- Let AI crawlers in. Check your
robots.txtdoesn’t block the crawlers used by AI search, such as OpenAI’sOAI-SearchBotandPerplexityBot, unless you have a reason to. - Fast, readable HTML. Content that only appears after heavy JavaScript runs may never be seen by some crawlers.
llms.txtis a proposed file that summarises a site for language models. It’s cheap to add, but adoption by the major assistants is still unclear, so treat it as optional, not a fix.
Publish what buyers actually ask
Look at the questions in your sales calls: “How long does a CRM migration take?”, “Should we hire a RevOps manager or an agency?”, “Clay or Apollo?”. Each is a prompt someone is typing into an assistant. A clear, honest article answering it, like our GTM engineer vs RevOps guide, is exactly what an assistant looks for.
How to measure AI visibility
You can’t see AI recommendations in Google Analytics, so measure them directly:
- Write 15 to 30 buyer prompts for your category, in the words customers use. Include broad prompts (“best GTM engineering agencies for SaaS”) and narrow ones (“who can build Clay enrichment workflows for HubSpot?”).
- Ask them in ChatGPT, Perplexity and Gemini every month, with web search on, and record whether you’re mentioned, how you’re described, and which pages are cited.
- Track competitors in the same answers. They show you which pages and third-party sources the assistants trust in your space.
- Ask new leads how they found you, and add “an AI assistant” as an option. It’s increasingly the honest answer.
Answers vary between runs, so look at trends over months, not single results.
What not to do
- Don’t publish hundreds of thin pages written only for machines. Assistants and search engines both get better at ignoring them, and human buyers notice.
- Don’t invent reviews, clients or numbers. Beyond the legal risk, a single false claim found by a buyer costs far more than an AI mention gains.
- Don’t treat it as separate from SEO. Most of what helps AI visibility (clear pages, real expertise, good structure, independent mentions) also helps traditional search.
The bottom line
AI assistants are becoming the first filter in many B2B buying journeys. The companies that get recommended are the ones that explain clearly what they do, prove it with specifics, and are talked about by others. None of that is a trick. It’s the work of being easy to understand and easy to trust, now with a new kind of reader.
Want to know whether ChatGPT and Perplexity recommend your company for the work you want to win? Get in touch and we’ll show you what the assistants say about you today.


