Is AI search worth investing in? Three studies published on October 21 and 22, 2026 answered no, and all three measured the one thing an AI answer is built to remove: the click. Semrush found AI Overviews on 88% of the commercial keywords it tracks, Marie Haynes found ChatGPT sending 0.62% of the sessions Google organic does, and Ahrefs said Google now alters 5% to 20% of the results served to the tools that measure both. Read together, they describe a market where the purchase decision has moved into the answer while the instruments that priced the old market are failing.

  • AI Overviews now sit on 88% of tracked US commercial keywords and 97% of their search volume, so consideration for software, banking and ecommerce purchases happens inside a Google-written answer.
  • ChatGPT sends 0.62% of the sessions Google organic does. That proves the referral is thin, not that the channel is worthless: a decision surface was never designed to send a click.
  • Google is feeding altered results to rank trackers, so the measurement layer the open web relied on for twenty years is degrading at the same time as the click.
  • The answer that decides a purchase is assembled from third-party review and comparison pages. That is publisher inventory with no click and no price, and it is the next media market.

What decision-stage AI Overviews are

Decision-stage AI Overviews are Google-generated summaries placed above the results for queries that signal a purchase is close: reviews, alternatives, best-X-for-Y, pricing, and brand-versus-brand comparisons. Semrush's October 21, 2026 study tracked 400 such keywords across banking, software and ecommerce and found coverage rose from 45.75% in September 2025 to 88% in August 2026. Review queries moved the most, from 30.14% to 89.04%.

These summaries draw on review sites, YouTube and comparison publishers at least as often as on the brand's own page. In Semrush's Mailchimp-versus-Klaviyo example, Klaviyo's page made the case for Klaviyo, Zapier made the case for Mailchimp, and a YouTube channel filled in the rest.

The ranked-link economy vs the answer economy

In the ranked-link economy, a commercial query produced a list, the list produced a click, and the click was the billing event for both the publisher's ad impression and the brand's cost per click. Three neutral instruments made that market legible to outsiders: the rank tracker, the analytics referrer and the conversion pixel.

In the answer economy, the same query produces a synthesized recommendation built from several sources, read in full and rarely clicked. The rank tracker is now served poisoned data, the referrer files AI Overview clicks under Google organic, and the pixel fires on a fraction of the journeys it used to see. The demand did not shrink. The unit that priced it did.

What PPC Land reported, and what we verified

PPC Land's October 23, 2026 digest, written by Luis Rijo, grouped the three findings and flagged the caveats on each: Semrush's panel has no click data and the company sells the tracking it recommends, Haynes' 100-site sample discloses no industry mix, and Ahrefs published no methodology. We checked every figure against the originals. Semrush's own post confirms 88% keyword-weighted and 97.08% volume-weighted coverage; Haynes' analysis, reported by Search Engine Land and MarTech on October 21, confirms 0.62% for ChatGPT and 0.73% for all LLMs; Ahrefs' October 22 product blog confirms the 5% to 20% range, and Nozzle's data puts its own exposure at about 8% with a 27% peak. No discrepancy found. One omission matters: Haynes did not measure how Google organic itself changed over the period, which is exactly where the AI Overview effect sits.

Why the 'GEO is a distraction' verdict reads the right data backwards

The strongest case for ignoring AI search in 2026

Marie Haynes' case is the strongest, and it is a budget argument rather than a traffic argument. Across 145 million sessions, Google organic drove 96.5% of conversions and 27 times the conversions of all LLMs combined, and ChatGPT referrals have declined since August 2025. SEO Works found organic delivering 39 times the traffic of AI assistants, and Kevin Indig found no reliable link between AI visibility and branded search across 12 sites. If budget follows conversions, the budget stays with Google.

Why referral share cannot price a surface built to end the click

The verdict fails because it measures a decision surface with a transport metric. Pew found in July 2025 that users click a link 8% of the time when a summary appears against 15% without one, and Ahrefs put the click reduction for top-ranked pages at 58% in February 2026. A surface that removes the click will always look worthless on a referral chart.

iPullRank's Mike King made the sharper point on October 22: AI Overview and AI Mode clicks carry a google.com referrer and land inside the Google organic bucket these studies use as their benchmark, while ChatGPT's weekly audience grew from 700 million to 1.2 billion over the same window in which its referrals fell. Usage rose and referral fell because the answer is doing the work the click used to do. The only way to be inside that work is to be inside the sources the answer is built from.

The Ahrefs disclosure closes the loop. The ranked-link economy had an independent referee for twenty years; when Google serves suspected scrapers swapped results and Search Console still reports no AI Overview query data, the market loses its neutral instrument at the moment the unit changes. A market whose measurement is controlled by its largest seller is not one a CMO can budget against from the outside.

What this means for brands and for publishers

For CMOs, media buyers and agencies: budget the answer, not the referral

Stop grading AI search on the sessions it sends and start grading it on the sources that decide the answer. For every decision-stage query in your category, map the pages the summary cites, treat them as media inventory, and fund presence there with a line separate from SEA and from GEO content. The 0.62% row in GA4 will never show you this, because it was never built to.

For publishers: comparison content is the raw material of nine in ten review answers

Your review and comparison pages now feed the answer for 89% of review queries and receive almost none of the resulting traffic. Lobbying for the Stealth Bot Prohibition Act, as the News/Media Alliance and a UK coalition including the FT and Condé Nast are doing, buys identity for the crawler, not a price for the citation. Price the future on being cited in commercial answers: that is inventory a brand will pay to sit beside, and it is worth more per citation than the click it replaced, because the decision is made there.

Three signals to watch before the 2027 budget cycle

Three moves show where this lands. Google now sells ads inside AI Overviews in 12 countries with no opt-out and no reporting column, per PPC News Feed's October 22 reference guide, so the platform is already monetizing the decision surface while denying buyers an instrument to measure it. Google's poisoning of rank-tracker data, which Search Engine Roundtable and SISTRIX read as aimed at AI companies and SerpApi, shows the answer layer's inputs are now treated as a defended asset. And CNN International is building for agent-led ad buying with first dollars targeted for Q1 2027, which means publishers are preparing to sell to machines before most brands have learned to buy around them.

Conclusion

Hold on to this: the purchase decision has moved into the AI answer, the answer is built from publisher pages, and the referral chart that says AI search is not worth investing in is measuring a unit the surface was designed to retire. Smalk AI is built for that gap: a Generative Engine Advertising network that places native ads for AI agents on the publisher pages AI engines read and cite, so brands buy presence where the recommendation forms and publishers earn media revenue for the citations that replaced their clicks. What to watch next: the first vendor to publish a citation-level price for a decision-stage query, because that number, not a referral ratio, is the one the 2027 media plan will be written against.