Ask an assistant to recommend software in your category. Not your brand, your category. The answer comes back in a paragraph with three or four companies named in it, and for most B2B companies I look at, the company is not one of them. If you run AI B2B SaaS checks like this for a living, the pattern is hard to unsee.
That is a different problem from ranking. On Google you can sit at position 11 and still be findable if someone scrolls. In an AI answer there is no position 11. You are named or you are absent, and the buyer never learns you existed.
First time I ran it properly was early September, an AI visibility report on the topic “Growth”, answers from Copilot and its partners. Piquiyo: zero citations. The most cited domain in the whole topic was a UK agency with under 10% share. I had never seen them in a pitch. The assistants had.
What is AI search visibility?
AI search visibility is whether your company gets named and cited when someone asks an AI assistant a question your product answers. It applies to ChatGPT, Perplexity, Claude, Gemini and Google’s AI Overviews, and it is measured by running real buyer questions and recording which companies appear.
It is not a ranking you can look up. There is no console for it, no position number, and no official figure for how often it happens. The only way to know is to ask and write down what comes back.
Does AI search matter yet for B2B SaaS?
It matters at the top of the funnel, where a buyer is building a shortlist and has not searched for a vendor by name yet. That is exactly the moment a category question gets asked, and it is the moment you are most replaceable.
Be careful with anyone quoting a precise share of B2B research now happening in AI assistants. Those numbers move fast and most of them are extrapolated. What you can measure is your own: your analytics will show referrals from assistant domains, and your sales team can be asked, on every call for a month, where the prospect first heard the category framed the way they framed it. That is first-party evidence and it is worth more than a benchmark.
How do I check whether we are cited?
Run a fixed set of buyer questions across the main assistants and record whether you appear. Same questions, same wording, once a month, so the results are comparable over time.
Build the set like this:
- Five category questions. “What software do B2B companies use to [job your product does]?” No brand names.
- Five problem questions. The exact wording a buyer uses on a first call when describing the pain. Take these from real call notes, not from your positioning doc.
- Five comparison questions. “X vs Y for [use case]”, using the competitors you actually lose to.
- Five buying questions. “How much does [category] cost?”, “What should I ask a [category] vendor?”, “Is [category] worth it for a 50 person company?”
- Five brand questions. “What is [your company]?”, “Is [your company] any good?”, “Who are [your company]’s competitors?” These tell you what the assistant already believes about you, which is sometimes worse than absence.
For each question, record three things: were you named, were you cited with a link, and which sources were cited instead. That third column is the useful one. It tells you which pages the assistants currently trust in your category, and those pages are your real competition.
Run it in a fresh session with no history, so you are seeing what a stranger sees rather than what your own past prompts have taught it about you.
Why are we absent when we rank on Google?
Usually because the page that ranks is written to be scrolled, not to be quoted. Assistants retrieve passages, not pages, so a section that only makes sense after reading the two above it is a section that cannot be lifted.
The pages that get cited tend to share a few plain properties. They answer the question in the first two sentences of a section. They state numbers in text rather than in a chart image. They name their sources in the sentence rather than in a footnote. And they commit to a claim instead of hedging, because a sentence stacked with qualifiers has nothing quotable in it.
There is no secret ranking factor in any of that. It is just what a passage looks like when it can be lifted out and still make sense on its own. Google says as much in its own guidance on AI features in Search: there is no special markup, the same fundamentals apply.
What changes about our content?
Less than the noise suggests, and the changes are structural rather than topical.
- Write sections that stand alone. If the first two sentences of a section do not make sense to someone who read nothing before them, rewrite them.
- Put the numbers in the text. Charts are for humans. Tables and sentences are for everything else.
- Answer the buying questions you have been avoiding. Pricing, contract terms, who you are wrong for. These are the questions people ask assistants precisely because vendors will not answer them on their own sites, so the assistant answers from a third party instead.
- Say who you are, plainly, once per page. What the company is, who it serves, in what market. Not a tagline. Your ideal customer profile should be legible from the page itself.
What does not change?
The work still has to be worth citing. An assistant will not name you because your headings are well formed. It names you because something on your page answered the question better than the alternatives it found.
So underneath the new vocabulary, the content problem is the one it always was. If everything you publish could have been published by four competitors with the logo swapped, no amount of structure will fix it. This is the same reason demand generation built on a real point of view compounds and generic content does not.
Not that we were absent. Who was present. Five domains I had never competed with for a client were holding a topic we do the actual work in. Rivals had dozens of pages on AI. We had zero. The assistants were not wrong about us. We had given them nothing to quote.
What I would do first
- Run the 25 questions. One afternoon. Record named, cited, and who was cited instead.
- Read the pages that got cited instead of yours. Not to copy them. To see what question they answered that you did not.
- Fix your three highest-intent pages for retrievability before writing anything new. Answer-first sections, numbers in text, sources named.
- Answer two buying questions you currently dodge. Pricing and fit are the usual pair.
- Rerun the set in 30 days. The point is the delta, not the first score.
Frequently asked questions
What is answer engine optimisation?
Answer engine optimisation is the practice of structuring content so AI assistants can retrieve and quote it when answering a question. In practice it means answer-first sections, facts in text rather than images, named sources, and direct claims. It overlaps heavily with good SEO and is not a separate channel.
How is AI search different from SEO?
SEO competes for a position in a list a person can scroll. AI search competes to be named inside a single generated answer, where there is no second page. The content work overlaps, but the failure mode is different: in SEO you rank low, in AI search you are simply absent.
Can I track AI search visibility in a tool?
Partly. Some tools now sample assistant answers for your brand, and your analytics will show referrals from assistant domains. Neither gives you a complete picture, because answers vary by session and by phrasing. A fixed question set you rerun monthly is more reliable than any single number.
How long does it take to start appearing?
There is no published answer to this and anyone quoting a specific timeline is guessing. What you can control is whether your pages are retrievable and worth quoting, and you can measure the change with your own question set.
Do we need new content or better content?
Better, almost every time. Fixing retrievability on pages that already rank is faster and cheaper than publishing new pages that nobody trusts yet. The same logic applies to B2B lead generation: fix the surfaces you already own before buying new traffic.
One thing to do this week
Open a fresh assistant session, ask the five category questions, and write down who gets named. If it is not you, you now know something about your next quarter that no dashboard was going to tell you.
That check is the first thing we run for new clients. If you would rather not do it yourself, we can run it with you.