Your brand can rank number one on Google and still be invisible inside ChatGPT, and for a fast-growing share of B2B buyers, only the second one now counts. To get cited by ChatGPT and other AI assistants, three things have to be true: the AI can read your pages, the wider web describes you consistently and credibly, and your own pages are structured so an answer can be lifted straight out of them. The biggest lever by far is the middle one. Roughly three-quarters of what decides your AI visibility happens off your own website, in the mentions, reviews, comparisons and community discussion that AI models learn to trust.
That is an uncomfortable reframe if you have spent years treating your website as the whole game. So this is a practical playbook, not a theory piece. It covers what "getting cited" actually means, how ChatGPT chooses what to pull in, and the exact steps a B2B marketing team can run, starting this week, to move from absent to recommended. If you want the broader context on putting AI to work across your marketing, we have covered that separately in AI in Marketing: A B2B SaaS Marketer's Field Guide.
Getting cited by ChatGPT means one of two related things, and they are worth separating because they need different work. A citation is a linked source: ChatGPT runs a web search, builds an answer, and links your page as evidence. A mention is your brand name appearing in the answer itself, often while ChatGPT cites someone else, a G2 listing, a Reddit thread, a comparison article, an industry roundup. Both drive demand. Only one points a link at your domain.
For B2B, the mention is usually the prize. When a buyer asks "best [your category] tool for a 20-person team" and your name comes back, the model is effectively vouching for you at the exact moment of decision. Buyer behaviour has already shifted to make this matter: the large majority of B2B buyers now use AI tools somewhere in their research, and a growing share open ChatGPT before they open Google. For some products this is no longer marginal. The bootstrapped form builder Tally has reported that ChatGPT became its single largest referral source.
You will see this work labelled several ways, generative engine optimisation (GEO), answer engine optimisation (AEO), or LLM optimisation. The names differ; the goal is identical. It is making your brand the source AI assistants reach for when they answer your buyers' questions.
ChatGPT does not have a single "citation algorithm" you can game. It sometimes answers from what it already learned in training, and sometimes searches the live web first, a process known as retrieval-augmented generation, meaning it pulls in current pages before writing the answer. When it searches, a useful mental model is: read your page, decide whether it trusts the source, then quote the part that answers the question. Three patterns matter for marketers.
First, it leans on consensus. AI models grow confident about a brand when many independent sources describe it in similar terms. If you only exist on your own site, the model treats your claims with healthy scepticism. If your site, Reddit, YouTube, G2 and a few industry publications all describe you the same way, it trusts you enough to name you.
Second, site authority counts more than page authority. Ahrefs' analysis of the pages ChatGPT cites most found that cited pages sit on domains with a median Domain Rating around 90, while the median rating of the individual page was only about 6. Being on a trusted domain matters more than that specific URL being heavily linked. The same analysis is sobering about where citations land: Wikipedia and homepages dominate, and only around a third of citations sit on page types you can realistically influence, educational and how-to content, reviews, news and blogs, and community threads. That influenceable third is where your effort belongs.
Third, ChatGPT is not just mirroring Google. Around 28% of its most-cited pages have no organic search visibility at all, so it discovers and trusts content on its own terms, weighting freshness, clarity and evidence. But there is a strong overlap worth exploiting: roughly 85% of ChatGPT-cited pages also rank for at least one Google keyword, and one Seer Interactive study found that 87% of ChatGPT search citations matched Bing's top results. Which leads to a detail most marketers miss: ChatGPT's web search runs on Bing's index, not Google's. If you are invisible in Bing, you cannot be cited there, however good your content is.
Because the strongest signals an AI uses to judge you are not on your website at all. When Ahrefs studied 75,000 brands, the single best predictor of appearing in AI answers was not backlinks, domain authority or how much you publish. It was branded web mentions, text written about your brand elsewhere, which correlated with AI visibility at 0.664, more than three times stronger than backlinks at 0.218. A more recent Ahrefs report puts YouTube mentions even higher, at around 0.737, the strongest single signal they measured across ChatGPT, Google AI Mode and AI Overviews. The effect is winner-takes-most: brands in the top quartile for web mentions earned up to ten times more AI citations than the next quartile down.
Other research lands in the same place. AirOps found brands are about 6.5 times more likely to be cited through external sources than through their own domain. A useful rule of thumb, from agency NAV43, is the 75/25 split: treat roughly 75% of your effort as off-site, earned mentions, reviews, community, original research others quote, and 25% as on-site, structuring your own pages to be easy to cite. Most brands do the exact opposite, pouring everything into the 25% they directly control.
This is the heart of it. We run this programme for B2B SaaS brands, and for SaaStrix itself, and the pattern is consistent: the teams that win AI visibility stop thinking like publishers and start thinking like the most-quoted expert in the room. The work below is sequenced around that.
Six moves, in order. The first tells you where you stand. The middle four do the work, framed around the three things that have to be true: AI can read you, trust you and quote you. The last keeps it alive.
You cannot improve what you are not sampling. Before changing anything, build a baseline.
One critical caveat shapes how you read this. SparkToro, with Gumshoe.ai, ran 2,961 brand-recommendation prompts across ChatGPT, Claude and Google's AI. The same list of brands came back less than 1% of the time, and the same list in the same order under 0.1%. A single screenshot proves nothing. Run each prompt several times, ideally dozens, and average it. Chase share of voice across many prompts, not a ranking on one.
This is the unglamorous prerequisite. If AI crawlers cannot reach or parse your pages, nothing else matters.
This is where the returns are largest, so give it the most time.
Now the on-site work. The goal is extractability: give the model a clean, self-contained passage it can lift.
Freshness is one of the most consistent and most neglected signals. Around three-quarters of ChatGPT's most-cited pages were updated within the previous 30 days, and SE Ranking found pages updated within three months averaged 6 citations against 3.6 for stale ones. Citations also decay: a page can quietly lose them by going out of date.
Set a refresh cadence. Treat your highest-value pages roughly every 90 days and key guides quarterly. Add a visible "last updated" date, and make the updates substantive, new data and new examples, not a changed timestamp.
Manual spot-checks are unreliable because answers shift constantly. Track signals over time instead: share of voice (how often you appear across your category prompts versus competitors), citation rate, the mix of owned versus third-party sources, sentiment and accuracy (is the model describing you correctly), and AI referral sessions and conversions in your analytics. For doing this at scale, the named sources point to Ahrefs Brand Radar (its Cited Domains and Cited Pages reports show exactly which third-party pages AI pulls from in your niche), Semrush's AI visibility tools, Moz, and Yoast SEO AI+. Re-run your prompt panel monthly, then double down on the themes that start earning citations.
Three honest caveats, because credibility is itself a citation lever.
It is volatile. The same SparkToro research shows lists changing on almost every run, and citation patterns shift when models update. Do not panic at fluctuation, and do not over-invest in any single tactic. Visibility across many prompts is far more stable than position within any one of them.
Reddit's role is contested. One analysis of 6.8 million citations found only about 2% came from forums, while other studies found heavy Reddit presence correlated with far more citations. The honest read is that it varies by industry and model. Participate genuinely, but do not bet everything on it.
And good SEO is still the foundation. With around 85% of AI-cited pages also ranking in Google, the fundamentals, crawlable pages, depth, authority and clarity, still do most of the heavy lifting. As Search Engine Land summarised it, "good SEO is good GEO". This is an extension of solid marketing, not a replacement for it.
You do not need to do all of this at once. In the first 30 days, measure: run your 15 to 20 buyer prompts across the AI assistants, log where you appear, and repeat for three competitors. In the next 30, fix the foundations: unblock the AI crawlers, claim and clean up your G2, Capterra and Trustpilot profiles with a review push, and pitch your way onto the two or three comparison articles AI already cites in your category. After that, compound: publish one piece of original data, seed consistent brand language across Reddit, LinkedIn and partner content, and set your refresh cadence.
If you would rather do this with other B2B marketers working through the same shift, getting found by AI is one of the practical workflows we go deep on inside the SaaStrix community, where AI Labs covers the exact prompt panels, tracking setup and templates step by step. The earlier you start, the more citation share you bank while the space is still uncontested.
Want practical, scannable AI content for B2B marketers delivered weekly? The SaaStrix newsletter has 21,000+ subscribers and is free to join at saastrix.uk/newsletter.
.svg){kind=icon}