Ask ChatGPT for a software recommendation, and it’ll give you three names. Maybe four. What ChatGPT won’t tell you is that it quietly passed over 40 other brands to get there – some of which rank on the first page of Google, some of which have better product reviews, some of which have been in the market for decades.
Nobody told those companies they weren’t going to make the cut. There was no notification, no penalty, no explanation. They just… didn’t appear.
So what’s actually happening? How does an AI model decide which brands are worth naming and which ones get silently ignored? It’s a question that matters enormously to anyone running an online business right now – and the answer is more mechanical than most people expect.
Signal One: What the Rest of the Internet Says About You
“ChatGPT doesn’t trust you. ChatGPT trusts everyone else talking about you.”
Start here, because this is the piece most businesses get wrong.
When someone asks ChatGPT for a recommendation – which CRM to use, or which accountant to hire or which hotel to book – the model isn’t pulling up your homepage and reading your About section. It isn’t impressed by your brand story or your mission statement. What it’s actually doing is looking outward, scanning the wider web to find out what the rest of the internet thinks about you.
Review platforms. G2, Capterra, Trustpilot, Tripadvisor – these get scraped. Industry roundup articles get scraped. Reddit threads get scraped. Blog posts that list the “top five tools” in any given category get scraped. The model is essentially asking a single question over and over ‘Does the broader web agree this brand is credible?’
If the answer is yes – if your name keeps appearing in the right conversations, on the right platforms, in the right context – you get cited. If you’re basically absent from those conversations, the model has no basis for trusting you, and it moves on.
This is a meaningful shift from how digital marketing has worked for the past 20 years. Traditional SEO was largely about backlinks – other websites linking to yours as a signal of authority. AI search is more about brand mentions, period. You don’t even necessarily need a link (a bold statement from a seasoned SEO). A Reddit thread where someone recommends your product by name, completely organically, is worth something. A listicle that includes you in its top five carries weight. A G2 review that describes your product in specific, accurate terms – that all feeds into the model’s picture of who you are.
The brands that consistently show up in AI answers have usually done a similar set of things: they’ve secured placement in the relevant listicles for their category, they’ve shown up in the forums where their buyers actually hang out, and they’ve kept their review profiles active. Not as a manipulation tactic, but because all of that activity creates a trail of third-party proof. The LLMs are looking for that proof. Give it to them.
Signal Two: Whether Your Website Can Actually Be Read by a Machine
(Your website might be invisible to a machine)
Here’s where things get a bit more technical – but stay with me, because this one catches a lot of digital marketers off guard.
You could have stellar reviews across every major platform. You could be mentioned in fifty industry roundups. And ChatGPT might still skip you, purely because of how your website is structured.
AI models aren’t browsing your site the way people do. They’re not scrolling, getting a feel for the design, noticing that your hero image is well-chosen. They’re scanning for structure. They’re trying to extract a clean, usable answer from your content – and if they can’t do that quickly, they move on to whoever they can extract from.
This is why the concept of “answer-first” formatting has started gaining traction among people paying attention to AI visibility. The idea is simple: every major section of your website, every landing page, every blog post, should lead with a direct answer to the question it’s designed to address – not warm up to it over five paragraphs. AI models have no patience for a slow build. They want the answer at the top, and the supporting context below.
Beyond writing style, there’s schema markup. This is a layer of structured code that essentially labels your content for machines – telling them what your organisation does, what category your product belongs to, who wrote a piece of content, and what questions it answers. Without it, the model has to guess. With it, extraction becomes reliable. The FAQ schema and organisation schema are worth implementing if you haven’t already. It’s not glamorous work, but it pays off.
And then there’s something called entity consistency – which sounds complicated but really just means this: pick a way to describe your company and your products, and use it identically everywhere. Your website, LinkedIn, Crunchbase, G2, everywhere. If your website says you’re a “B2B revenue intelligence platform” and your LinkedIn says you’re a “sales analytics tool”, and your G2 profile says something else entirely, you’re sending conflicting signals. The model genuinely can’t reconcile these into a clean picture of what you are. So it defaults to whoever has their story straight.
Signal Three: Your Standing in Traditional Search Still Matters
“Google isn’t dead. It’s actually feeding the beast.”
This one tends to surprise people who’ve drunk the “AI is replacing search” Kool-Aid.
ChatGPT – especially when handling commercial queries, the “what’s the best X for Y” questions – frequently doesn’t rely solely on its training data. It pulls live results from the web, often from Google or Bing, to build or supplement its answers (ChatGPT definitely uses Google to search the web). There are reports that OpenAI has leaned on services that scrape live Google results directly into the model’s responses. What this means practically is that if you rank well in traditional search, you have a structural advantage in AI recommendations too. Google has already decided you’re relevant and trustworthy – and the AI model picks up on that signal.
None of this means you can go back to keyword stuffing and call it a day. The content still needs to be structured for machine readability, answer-first, with schema in place. But it does mean that businesses writing off traditional SEO in favour of chasing pure “AI visibility” are making a mistake. The two things reinforce each other. Google authority bleeds into AI authority, and the brands treating these as separate games are playing with a handicap.
A few practical things follow from this. Keep your content updated – both Google and AI models weigh freshness, so content that hasn’t been touched in two years is a liability. Maintain your Google Business profile and Knowledge Panel. And stop looking at your Google rankings and your AI visibility as separate dashboards tracking separate realities. They’re not. They’re feeding the same ecosystem.
One more layer worth understanding here: there’s a category of authority signals that sits above your own content entirely. Industry awards. Wikipedia entries. Appearances in major comparison pieces in trade publications. These matter because they represent independent, credible sources saying your brand is real and worth knowing about. The model treats those external validations differently from self-published content – they’re harder to fake, so they carry more weight.
What You Should Actually Do With This Information
Understanding the theory is one thing. Here’s the part that actually changes your situation.
Figure out where AI is finding your competitors, not you.
This is the starting point. Every category has its own ecosystem of trusted sources – the specific subreddits buyers use, the listicle sites that keep getting cited, the review platforms that dominate that vertical. The mix is different for SaaS than for hospitality and for professional services. You need to map the sources that AI is actually pulling from in your category, then ask honestly, “is my brand appearing in those places?” If not, that’s where you start building.
Test your own content for extractability.
Go to your most important pages – not all of them, just the two or three that should be doing the most work – and ask a single question: if a machine scanned this page right now, could it pull a clean, direct answer to the core question this page is supposed to address? Run that test honestly. If the answer is no, restructure. You don’t have to rebuild the whole site. A handful of pages, made genuinely clear and answer-first, will move the needle.
Set up actual tracking.
This is the step most people skip, and it’s the reason they don’t know whether anything they do is working. AI answers aren’t static – they shift as models update, as new content gets indexed, as competitors gain or lose mentions across the web. You need a baseline. You need to know which prompts you’re showing up in and which you’re invisible in. In Google Analytics, you can create a custom channel group that captures referral traffic from AI tools – filter by sources like chatgpt.com, perplexity.ai, and similar. This connects your AI visibility to actual sessions and conversions, rather than leaving it as an abstract exercise.
Note: If the C-suite in your company doesn’t know how to use GA4, then track AI traffic in Google Looker Studio.
The Uncomfortable Truth About How This Works
ChatGPT is not mysterious. It’s not making aesthetic judgments or playing favourites. It’s doing something fairly mechanical: finding the brands that appear most credible, most consistently referenced, and most clearly understood across the data it has access to. The brands that win are the ones that have made it easy for the model to trust them – through third-party mentions, through clean and structured content, through the authority that comes from ranking well in traditional search.
(The three signals that govern those choices – third-party trust, machine readability, and search authority – are not arcane technical requirements. They are, ultimately, expressions of the same thing: demonstrable, verifiable credibility. Build your presence where your buyers talk. Make your content easy for machines to understand. Maintain your authority in traditional search. Do all three consistently, and you give AI every reason to name you.)
The brands that lose aren’t necessarily worse. They’re just harder to read, harder to find in the right places, or inconsistent in how they present themselves to the world.
That’s actually good news. It means this is fixable. It’s not about budget or luck or being first to market. It’s about doing the unglamorous work of being consistently present, clearly structured, and externally validated. None of it is technically difficult. Most of it is just… not yet done.
The window to get ahead of this is still open. It won’t be forever.
Leave a Reply