How AI engines are reshaping where buyers find brands, and why trademark owners need to start measuring their AI visibility now.
Last updated: June 4, 2026
Key takeaways
- ChatGPT crossed 700 million weekly users in 2025, and Google's AI Overviews now answer informational queries directly without sending traffic to publishers.
- Brand navigational search ("type Nike into Google") is being replaced by AI-mediated brand questions ("ask ChatGPT what Nike is doing in sustainability").
- Research from HubSpot's 2026 study of 10,000 URLs across 8 AI engines shows that 55% of AI citations happen at the evaluation stage, when buyers are about to decide.
- For trademark owners, this means brand visibility is now AI-mediated. If your brand isn't cited in the answer, you've lost a buyer you'll never see in your analytics.
- The fix is not better SEO. It's a different category of measurement called AEO (Answer Engine Optimization), which tracks whether your brand appears in AI responses for the prompts your buyers actually type.
What's actually happening to search traffic?
The data is brutal. According to Search Engine Journal's 2025 analysis, queries that trigger Google's AI Overviews see click-through rates drop by 25 to 35 percent on average, and informational queries can lose nearly 50 percent of their organic clicks. Google has been transparent that AI Overviews are designed to answer the question on the page, not send the user elsewhere.
Meanwhile, OpenAI confirmed in mid-2025 that ChatGPT serves over 700 million weekly active users, up from around 200 million the year before. Perplexity reports more than 20 million monthly users and growing. Gemini is being baked into the Android system experience. Microsoft Copilot ships with Windows.
The buyer journey that used to start in a Google search box is increasingly starting in an AI chat box. And the AI chat box has a different relationship with brand visibility.
Why is this different from a normal search engine?
A traditional search engine returns a list of links. Even if your page is at position 5, the user might still click it. They see your brand in the SERP. They form an impression. Some fraction of them visit your site, and your analytics records the click.
An AI engine returns an answer. If your brand isn't named in that answer, you don't exist for that query. There is no list of "also-rans". There is no impression. There is no click. The buyer reads the answer, makes a decision, and moves on. You only find out you were skipped when the deal goes to a competitor you've never heard of.
This is why HubSpot's 2026 research on 10,000 URLs across 8 AI engines was so striking: they found that backlink count has near-zero correlation with AI citations (between -0.04 and +0.06, depending on the engine). The signals that earned you traffic from Google barely matter to ChatGPT, Perplexity, or Gemini. The game has changed.
What does "AI visibility" actually mean?
AI visibility, sometimes called Answer Engine Optimization or AEO, measures three things:
- Mention rate. When a buyer types a prompt in your category, does the AI engine include your brand in the answer?
- Position and sentiment. If you're mentioned, are you the top recommendation or a footnote? Does the AI describe you favourably or with a caveat?
- Citation share. Which URLs does the AI cite as evidence? Are those URLs you control, sites you've earned coverage on, or sources you've never heard of?
The buyer never sees the prompt you'd care about. They see the answer. So the measurement question is: are you in the answer for the prompts your buyers ask?
Why does the buyer's prompt matter more than their search query?
A search query is short and ambiguous. "best CRM" returns ten blue links and the buyer picks. A buyer who types "best CRM for a 12-person services agency selling to legal" gets a list of three to five recommendations, with reasoning, written in a way that looks like advice from a colleague. The buyer reads the recommendations and shortlists from them. There is no second page.
This is why several independent analyses have converged on a striking funnel finding. HubSpot's research breaks AI citations across the four buyer journey stages as follows:
- Solution education ("what is X"): 12% of citations
- Problem exploration ("why does Y happen"): 14%
- Solution comparison ("X vs Y vs Z"): 19%
- Solution evaluation ("best X for our situation"): 55%
More than half of AI citations happen at the evaluation stage. This is the opposite of the SEO traffic distribution you're used to, where most clicks come from top-funnel awareness queries. AI engines disproportionately get used at the bottom of the funnel, when the buyer is choosing.
What does this mean specifically for trademark owners?
Trademarks exist to give brand owners a defendable, distinctive identifier. The whole point of a registered mark is that buyers can recognise your brand and distinguish it from the competition. As USPTO puts it, a trademark "identifies the source of the goods and distinguishes them from others".
That recognition step is now AI-mediated. A buyer searching for legal help for trademark filing in 2020 typed your brand into Google, found your website, and recognised the mark you'd spent years registering. In 2026 they type "best service for filing a trademark for a SaaS startup" into ChatGPT and read three recommended providers. If you're not one of the three, the years of brand investment, the USPTO registrations across multiple classes, the consistent visual identity, none of it converted that buyer.
You can look up which classes a competitor has filed in via the owner profiles on WikiTrademarks. You can see how many filings a brand has made and where they've expanded via the most valuable brands index. What you can't see, without a tool dedicated to it, is whether AI engines are talking about that competitor when buyers ask the buying question. That gap is the AEO measurement gap.
What about brand defence and infringement?
There's a second angle that matters for IP holders. If AI engines start describing a competitor with language that's confusingly similar to your registered mark, or if they recommend a counterfeit version of your product alongside your real one, you have a brand dilution problem that doesn't show up in any traditional monitoring tool.
The International Trademark Association has been warning since 2024 that AI-generated content creates new infringement surface area. The WIPO position paper on AI and IP raised similar concerns. Tracking AI mentions of your brand isn't only a growth play. It's a defensive monitoring play that complements the TTAB monitoring you may already be doing for traditional opposition and cancellation proceedings.
How do you measure AI visibility?
Three honest options:
1. Do it manually. Pick 10 prompts your buyers care about, type each one into chatgpt.com, perplexity.ai, and gemini.google.com weekly, and log whether your brand appears. This works for one brand at small scale. It does not scale to a multi-brand portfolio, and humans miss subtle position and sentiment shifts.
2. Use an API-based AEO tool. Tools like Profound, Otterly, Peec, and HubSpot's new AEO module hit the OpenAI, Anthropic, and Google APIs and report what comes back. This scales. The catch is that the API and the web app return different answers because the web app uses retrieval, memory, plugins, and a different system prompt. If you're optimising for what your buyer actually sees, API data is the wrong source.
3. Use a tool that captures real chatgpt.com responses. Our own AEO Tracker uses a Chrome extension running on a dedicated machine to capture real chatgpt.com responses, which matches what your buyers actually see. This is more expensive to run, which is why it's a paid product, but the data fidelity is meaningfully better.
What's the right first move?
Whichever option you pick, the first move is the same: pick your 10 to 20 highest-intent prompts. Not "what is a trademark" (top-funnel, low conversion). Prompts that look like the questions your highest-value buyers ask when they're choosing. For a trademark filing service that might be "best service for filing a trademark for a clothing brand selling on Shopify". For a brand monitoring tool it might be "how do I track if my brand is being mentioned in news coverage". Specific. Bottom-funnel. Decision-stage.
Then start running those prompts weekly and tracking the answers. Week over week, look for: are you being added or dropped? Has sentiment shifted? Have new competitors appeared? Have the cited sources changed? Each of those is a signal you can act on.
Frequently asked questions
Is AI search actually replacing Google search?
Not yet, but it is taking a measurable share, particularly at the evaluation stage of the funnel. Google's own data from 2025 showed AI Overviews now appear on roughly 13 percent of US searches and reduce CTR on those queries by 25 to 35 percent. ChatGPT's 700 million weekly users represents a separate, additional pool of buyer attention that didn't exist three years ago.
What is AEO (Answer Engine Optimization)?
AEO is the practice of optimising for AI engines that return answers rather than lists of links. Where SEO targets blue-link rankings, AEO targets being named in the answer ChatGPT, Perplexity, Gemini, and similar engines generate. It's measured by mention rate, position within the answer, sentiment, and which URLs the engine cites.
Do backlinks help with AI citations?
No, not really. HubSpot's 2026 study of 10,000 URLs across 8 AI engines found backlink count has a correlation of between -0.04 and +0.06 with AI citations. The signals that move the needle are structural (question-format headings, FAQ schema, visible "last updated" dates) and contextual (being cited on Reddit, LinkedIn, YouTube, and trade press, where AI engines disproportionately pull from for high-intent queries).
How is "AI visibility" different from social listening or brand monitoring?
Social listening tracks what humans say about your brand on social platforms. AI visibility tracks what AI engines say about your brand to the humans who ask them. The difference matters: a buyer in 2026 might never post about you on social, but they will absolutely ask ChatGPT about your category before they buy.
Why does it matter whether the data comes from the API or chatgpt.com?
Because they return different answers. The chatgpt.com web app uses retrieval, memory of prior conversations, plugins, browsing, and a different system prompt than the raw API. Your buyer is using the web app. If your tracking tool is hitting the API, it's measuring a different product to the one your buyer experiences.
Where should I start if I want to track this for my brand?
Pick 10 to 20 evaluation-stage prompts your buyers actually type, then run them weekly through chatgpt.com and log whether your brand appears, where in the answer, and which sources are cited. If you want this automated with real chatgpt.com data rather than API data, our AEO Tracker handles the weekly capture and gives you a diff each week. The waitlist is open while we onboard the first cohort.
Want to see whether ChatGPT mentions your brand for the prompts that matter? Join the AEO Tracker waitlist. We're onboarding a handful of brands at a time. Pricing starts at $79 per brand per month, cancel anytime.