Key takeaways
- Purchase intent is migrating into ChatGPT, where standard analytics can't see it.
- Track two separate metrics: whether AI cites your page and whether it actually names your brand.
- Pages that answer the question in their first lines are more likely to be cited and quoted.
- Run scheduled high-intent prompt tests to measure your 'share of answer' versus competitors.
- Keep ranking data clean and watch personalised AI ads — commerce signals will blur fast.
Buying decisions are increasingly starting — and sometimes finishing — inside ChatGPT. The problem for marketers is that this shift leaves almost no trace in the data you normally rely on. There's no tidy referral string, no keyword report, no clean attribution path. Yet the signals are real, and the brands that learn to read them will own the next wave of AI-driven commerce. This is a practical look at where purchase intent now lives, what the latest industry data tells us, and how to actually measure it.
Why your existing data is going blind
Classic SEO measurement assumes a search, a click, and a landing page. AI assistants break that chain. A user asks ChatGPT to compare two products, receives a synthesised answer, and may never click through. Recent OpenAI usage data shows that "looking things up" has grown to roughly rival writing as a share of consumer messages — meaning a large and rising slice of ChatGPT activity is informational and research-driven, exactly the behaviour that precedes a purchase.
That matters because the research stage is where brand consideration is won or lost. If your brand is absent from the model's answer, you're not even in the running — and you'll see none of it in your analytics.
The uncomfortable finding: being cited isn't being named
Here's where the data gets sharp. A comparison-prompt study found that some brands were rarely named by logged-out ChatGPT even when the pages ChatGPT cited explicitly mentioned them. In other words, your content can be in the model's source set and still never make it into the spoken answer the user actually reads.
This splits AI visibility into two distinct metrics every brand should now track:
- Citation presence — does the model pull your page as a source?
- Named mention — does the model actually say your brand name in the answer?
The gap between the two is a measurable weakness. Closing it is the new SEO.
What gets your brand into the answer

Complementary data on AI-cited pages points to a clear pattern: pages that AI cited tended to answer the question in their opening lines. The gap was smaller against pages that were retrieved but not cited — but the direction is unambiguous. Front-loading a direct, unambiguous answer increases your odds of being both pulled and quoted.
For purchase-intent queries, that means your product and comparison pages need to state, early and plainly: what the product is, who it's for, what it costs, and how it compares. Burying that under brand storytelling is now a quantifiable liability.
Building a purchase-intent measurement loop
You can't manage what you don't measure, and you can't measure AI buying signals with yesterday's dashboards. Here's the loop we build for clients:
- Prompt testing: Run a fixed battery of high-intent prompts ("best X for Y", "X vs Z", "is X worth it") on a schedule, logged-out and logged-in, and record whether your brand is cited, named, and recommended.
- Share of answer: Track the percentage of intent prompts where you appear versus named competitors. This is your AI equivalent of share of voice.
- Source-to-mention ratio: Measure how often your cited pages actually convert into a spoken mention — and diagnose the content gaps when they don't.
- Downstream traffic tagging: Watch for direct and branded-search lifts that correlate with AI visibility, since assistants often drive un-attributed demand.
Don't abandon the fundamentals

AI visibility is built on the same foundation as search visibility, so the quality of your ranking data still matters. Semrush's work on "poisoned" SERPs — validating ranking signals against manipulated results — is a useful reminder that noisy data leads to bad decisions in both worlds. Likewise, when choosing a tooling tier, the practical questions are the same ones that separate a plan like Semrush Starter from Pro+: how much historical data, how many tracked keywords, and how deep your domain comparisons go. Pick the tier that lets you monitor competitors at the resolution your market demands.
And keep one eye on the platforms' own direction. Google is already testing AI Mode ads informed by connected Gmail and Photos data, with privacy policy exceptions to allow it. The trajectory is clear: assistants are becoming personalised commerce surfaces where paid and organic buying signals will blur. The brands measuring intent now will be ready to buy, optimise, and win there first.
The bottom line for decision-makers
Purchase intent hasn't disappeared — it has migrated to a channel most analytics stacks can't see. Treat AI visibility as a first-class metric, separate citation from mention, answer questions fast and plainly, and keep your ranking data clean. That's how you turn invisible AI conversations into measurable pipeline.
Sources
- Semrush Starter vs. Pro+: Which plan is right for you? — Semrush Blog
- ChatGPT Rarely Named Some Brands Even When Sources Mentioned Them via @sejournal, @MattGSouthern — Search Engine Journal
- Poisoned SERPs: An update on Google ranking data in Semrush — Semrush Blog
- Looking Things Up Now Rivals Writing In ChatGPT, OpenAI Data Shows via @sejournal, @MattGSouthern — Search Engine Journal
- Google Privacy Update Lets Connected Apps Shape AI Mode Ads via @sejournal, @brookeosmundson — Search Engine Journal
- Pages AI Cited Got Straight To The Answer More Often, Data Shows via @sejournal, @MattGSouthern — Search Engine Journal