Ask ChatGPT which brand of running shoe to buy. The answer will name three or four. Whichever brands are in the answer have entered the consumer's shortlist before any website has been visited, any ad has been served, any retargeting pixel has fired. Whichever brands are not in the answer have lost the consumer in a layer the conventional marketing stack does not measure.
The scale of the layer is not theoretical. ChatGPT carries roughly 900 million weekly active users. AI Overviews appear in approximately 48% of Google queries as of early 2026. Around 93% of AI search sessions end without a single website click. Inside that volume, a brand is named or it is not, the tone is favourable or it is not, the competitor set is your competitor set or it is not, and the dashboards on the marketing floor detect none of it.
The metric replacing search rank as the primary measure of digital brand presence is mention share inside synthesised answers. It is measurable. It cannot be deferred for another budget cycle. And the brands that have started measuring it have produced case-study data the rest of the category has not had access to yet.
The Shift In One Paragraph
The funnel did not collapse. It moved upstream.
For the better part of fifteen years, the brand discovery story ran through a search results page. Brands optimised to rank. Agencies optimised to drive traffic. Performance teams optimised to convert. Each layer had a metric and each metric had a dashboard. That model is now incomplete. According to Similarweb's January 2026 panel data, AI tools are used by 35% of consumers at the discovery stage, against 13.6% for traditional search. The discovery moment, where shortlists are built and competitors are mentally eliminated, increasingly happens inside an AI response a brand has no editorial control over.
The traffic numbers obscure the scale of the change. AI referral traffic is small relative to organic search and growing slowly, at roughly 1% month on month across most industries. The low referral volume has led some teams to dismiss the channel. The dismissal is misreading the data. Users arriving from ChatGPT spend 15 minutes on site on average, against 8 for Google referrals, generate 12 pageviews against 9, and convert to transactional sites at 7% against 5% from Google. Some sectors report LLM visitor conversion rates as high as 15.9% from ChatGPT, more than nine times the typical organic search rate. The volume is small. The intent is sharp. The decisions are being made before the click.
Which means the click is no longer the moment of interest. The mention is.
Why Good SEO Does Not Automatically Produce Good AI Visibility
The playbook that worked for ten years partly inverts.
Traditional SEO optimises for ranking against a query. AI visibility optimises for being cited as a source or named as a brand inside a synthesised answer. The two disciplines overlap. The underlying mechanics are different enough that strong SEO does not reliably produce strong AI visibility.
Each platform has its own editorial personality. ChatGPT cites sources roughly 87% of the time but names specific brands in only about 20% of answers. Gemini does the inverse: names brands in around 84% of responses but generates a citation link in only about 21%. AI Mode and AI Overviews sit somewhere between. The same brand can appear strongly in one platform and almost not at all in another, and a single-platform measurement strategy will systematically misrepresent the brand's actual AI presence.
Brand mentions in trusted third-party sources are now the strongest correlate of AI visibility. Ahrefs research from December 2025 identified YouTube mentions and branded web mentions across earned media as the top factors. A 2025 Stacker study found that distributing content to a wide range of publications increased AI citations by up to 325% relative to publishing on a single owned domain. The SEO instinct to consolidate authority on a single domain is partially in tension with the GEO instinct to distribute brand presence across many trusted sources.
Sentiment varies enormously between platforms. One published study covering 6,447 brand mentions found that Perplexity returned positive sentiment in 76.9% of mentions, while the same brand set scored on ChatGPT at roughly one fifteenth of that rate. A 14.8x gap on the same brand, same prompt set, different platform. Tracking sentiment on one platform tells a misleading story about how the brand is described across the actual surface where consumers are listening.
AI recommendations are also less stable than search rankings. SparkToro's research found that asking ChatGPT or Google AI the same question 100 times returns the same brand list less than 1% of the time. Source citations change 40% to 60% month over month on AI Mode and ChatGPT. A position in an AI answer is a probability distribution, not a rank. The measurement has to reflect that.
What Brands Are Not Measuring Yet
Most brand measurement infrastructure is not built to handle the above. SEO tools rank pages. Brand monitoring tools count social mentions. Neither captures what the AI says when a consumer asks for a recommendation, in what tone, against which competitors, on which platform, with what stability over time. The gap is structural.
A 2026 Brandi AI prediction estimated that 92% of marketers plan to optimise for AI search while only around 41% currently do. The Conductor 2026 survey found that 32% of digital marketing leaders now rank generative engine optimisation as their top priority for the year. The intent is widespread. The execution is concentrated in a handful of early-mover brands. The GEO measurement market itself is projected to grow from $848 million in 2025 to over $33 billion by 2034, a 50.5% compound annual growth rate. That growth rate is the tell. Spending follows the brands willing to act on the measurement gap before their competitors do.
What Measurement Looks Like When It Works
The brands taking this seriously have moved past general dashboarding and onto specific, repeated, benchmarked measurement against named competitors. The approach Alkimi has built into its LLM Search Lift Study, which sits inside the standard minimum spend on every Alkimi campaign rather than being sold separately, is one example of how that measurement is being structured.
The study tracks four things.
How often a brand appears in AI-generated responses to a defined set of category prompts. Not just whether the brand appears, but at what frequency relative to competitor brands across a stable prompt set. The frequency is the visibility share.
In what tone the brand is described. Sentiment scoring across mentions, with attention to descriptive language (positive, neutral, dismissive, comparative) and the specific qualifiers that recur (such as "premium", "affordable", "controversial"). The tone is the editorial position the brand holds inside the AI's worldview.
Against which competitors. Benchmarked against up to five named rivals in the same category, on the same prompts, across the same platforms. The right competitor set is specific to the brand and the category, not a generic top-ten list.
How the picture changes pre- and post-campaign. Visibility share, sentiment and competitive position measured before a campaign goes live and again afterwards, on the same prompts. The shift is the lift attributable to the campaign. Without the pre/post comparison, any AI visibility number is a snapshot with no causal claim attached to it.
The specific instrument is not the point. Similarweb's AI Search Intelligence, BrightEdge's Generative Parser, Profound and Peec AI are all building in the category, each with strengths worth comparing on a real brief. The brands measuring this at all are about to hold a structural information advantage over the brands that are not. The brands measuring it against the campaign window will hold it earlier.
What the Campaign Data Has Already Shown
Two patterns are now visible in the campaign data, including Alkimi's own.
Visibility share moves faster than the quarterly-tracking instinct expects when earned media and brand activity are concentrated within a narrow window. AI models update their effective representation of a brand based on the corpus of recent content they are grounded on. A brand that has been quiet for six months and then runs a high-volume campaign with coordinated press coverage shifts its representation in that corpus measurably within the campaign window. AI visibility lift is not a separate workstream from a media calendar. It is an output of one.
Sentiment is harder to move than visibility. A brand can buy its way into more mentions in a quarter. It cannot buy its way out of a negative editorial framing in the same window. Sentiment tracks closely to the reputational signals the AI is reading: independent review coverage, third-party rankings, customer-voice content. An AI visibility strategy is part media planning, part PR strategy, part product-quality strategy. The brands integrating all three move faster than the brands treating this as a content optimisation problem.
The Polestar connected-TV activation Alkimi ran is a usable illustration. The campaign delivered a 34% increase in sales intent, a 24% lift in brand association with "100% electric", 99% viewability and a 96% video completion rate, all measured by Nielsen rather than self-reported. None of those are AI visibility metrics. They are exactly the kind of credible third-party signals that subsequently feed how AI models describe the brand. AI visibility is downstream of brand evidence. The campaign was upstream of all of it.
The Quarter's Work
Three actions are defensible regardless of which measurement vendor a brand eventually chooses.
Run a baseline AI visibility audit across ChatGPT, Gemini, Perplexity and Google AI Mode using the same prompt set for all four, against the same three to five named competitors. The output is a single matrix: visibility share, sentiment and competitive position per platform. Without the baseline, any future claim about AI lift is unmeasurable. Most brands have never produced the matrix once.
Repeat the baseline quarterly. The cost of running it twice a year is trivial. The cost of being one of the brands measuring lift in 2027 against no prior measurement is structural.
Integrate the matrix into the existing brand health stack rather than running it as a separate dashboard nobody reads. AI mentions sit alongside brand uplift, share of voice, attribution and sales lift. They are not a competing measure. They are a leading one. A brand's share of voice in 2024 told you what people were saying. A brand's mention share in ChatGPT in 2026 tells you what people are about to buy.
The shortlist is being built whether the brand is watching or not. The brands watching will know what to do about it. The brands measuring it against the campaign window will know what to do about it next quarter, instead of next year.
---
Related reading from Alkimi Research
What Your Clients Are Actually Asking When They Ask About Agentic. The parallel conversation happening inside agency planning rooms.
Is your Brand ready for Agentic Advertising? The readiness framework connecting AI visibility to the broader brand infrastructure question.
Benchmark your AI visibility
The LLM Search Lift Study is included within minimum spend on every Alkimi campaign. Benchmarking against up to five named competitors, pre- and post-campaign, across ChatGPT, Gemini, Perplexity and AI Mode. Reach lauren@alkimi.org or marco@alkimi.org for details.