Most of what gets written about AI search is vibes. So we pulled the 2026 studies that actually have numbers behind them, the field experiments, the large-sample crawls, the buyer surveys, and lined them up. Here is what the data says is true right now, what quietly stopped working, and what a business should do about it. We run this same playbook for clients, so the bias here is toward what moves the result, not what sounds good in a pitch.
The short version: AI answers now sit on top of about half of all searches, they are measurably eating clicks, and the thing that gets you cited inside them is not what got you ranked in Google. If your strategy still assumes "rank #1 and the traffic follows," it is already a year out of date.
Depending on the tracker, AI Overviews appear on 42% to 48% of Google queries in early 2026. The spread by query type matters more than the headline number. Comparison queries ("X vs Y") trigger an AI Overview about 95% of the time. Question-format queries, about 86%. And the one that should get every local business's attention: "near me" queries trigger an AI Overview about 77% of the time, far higher than anyone expected a year ago. AI answers are rarest on transactional queries, where Google still protects the click path to a purchase.
Drop in organic clicks when an AI Overview is present, measured by a randomized field experiment from Carnegie Mellon and the Indian School of Business (1,065 participants, published April 2026). Zero-click searches rose from 54% to 72% on those queries.
That is the cleanest number in the space, because it came from a real randomized experiment, not a correlation. When the researchers hid AI Overviews for one group, that group clicked out to the open web 38% more often. The same study found user satisfaction was unchanged whether the AI answer was shown or hidden, which directly contradicts the "people are happier with less clicking" line.
An Ahrefs study of 300,000 keywords found the click-through rate on the number-one organic result fell about 58% on keywords that trigger an AI Overview. The click did not move down the page. It disappeared. This is why "we still rank #1" is no longer the same as "we still get the traffic."
For twenty years, ranking in Google's top 10 was the whole job. In 2026 it is the entry fee for one game and close to irrelevant for the other. Moz found that 88% of Google AI Mode citations do not come from the organic top 10. A separate analysis found 37% of AI-cited domains do not rank in traditional search at all. And Google's own two AI surfaces, AI Overviews and the newer AI Mode, cite the same URL only about 14% of the time.
Read that again. The same company, answering the same question two slightly different ways, pulls almost entirely different sources. These are separate retrieval systems with separate logic. Optimizing for one does not get you the other, and optimizing for classic blue-link rank gets you neither by default.
None of this means SEO is dead. The technical foundation that makes you rank, crawlable HTML, fast pages, clean structure, real content, is also what makes you eligible to be cited. It is necessary. It is just no longer sufficient, and treating it as the finish line is the mistake.
This is the finding that reframes the whole discipline. Muck Rack analyzed 25 million links across the major engines and found about 84% of AI citations come from earned media, journalism, research, and reference sources, not from a brand's own website. AI systems treat third-party editorial coverage as the proxy for whether a brand is real and worth naming.
Brand mentions across the web correlate about three times more strongly with AI Overview visibility than backlinks do (Ahrefs, 75,000 brands). E-E-A-T signals predicted citation probability far better than domain authority, which barely moved the needle.
So the old link-building reflex is weak here. A backlink helps, but an unlinked mention in a publication an engine already trusts can help more, because the engine learned the brand from that source in the first place. The practical translation: the highest-value work is getting referenced in the places AI reads, then making sure your own site is clean and current enough to be the page it links when it does name you. We go deeper on this in our breakdown of what a real AEO agency actually does.
Treating "AI search" as one thing is a mistake. The major engines pull from visibly different places, and the 2026 source studies make the differences usable:
The takeaways are concrete. Keep key pages current, because freshness is one of the few on-page levers with a measured multiplier. Put real, quotable statistics on the page, because they get pulled. And do not ignore where your category actually gets discussed, because that is where Perplexity and ChatGPT are reading.
Being the frontier means saying the unglamorous part out loud. Several tactics that got sold hard in 2024 and 2025 do not hold up against 2026 evidence:
One more honest note on the platforms: Google now publicly calls AEO and GEO "still SEO" and resists the idea that it is a separate discipline. Microsoft went the other way, formalizing GEO in its guidelines and shipping the first real first-party AI-citation dashboard inside Bing Webmaster Tools. The vocabulary is still being argued over. The underlying work, structure, authority, and earned coverage, is not.
If the on-page numbers feel abstract, the buyer-behavior numbers are not. Forrester's 2026 survey of about 18,000 business buyers found twice as many named generative AI as their most meaningful research source than any other option, ahead of vendor websites and sales reps. 94% reported using AI somewhere in the purchase. More than half now build an internal shortlist before they ever touch a vendor's site. Forrester's own summary is blunt: the model of driving traffic to your site to then nurture them is getting much less effective, because the research happened somewhere you were not invited.
Local is where this gets sharp for most Calgary businesses. The SOCi Local Visibility Index found only about 1.2% of business locations get recommended in ChatGPT, against 35.9% that show up in Google's local 3-pack. AI is dramatically more selective about which local businesses it will name. Worse, when AI does describe a local business, it matches the Google Business Profile data only about 68% of the time. It is confidently wrong about a third of the time. And about 45% of consumers now use AI for local business recommendations.
We have local data of our own. When we ran 50 ChatGPT queries about Calgary marketing agencies, nine firms received nearly all the citations, and three signals predicted who got named almost perfectly. And when we audited 152 Calgary small business websites for AI readiness in our State of AEO in Calgary study, zero scored Strong. The demand is here, the engines are answering, and almost nobody local is set up to be the answer. That gap is the opportunity, and it closes as competitors wake up.
Strip it down and the playbook is short:
The brands that treat this as a passing trend will spend 2026 watching their traffic erode without a clear cause. The ones that treat it as the new shape of search will be the names the answer gives. In a city where almost no one is ready, being early is the entire advantage.
Depending on the tracker, AI Overviews appear on about 42% to 48% of Google queries in early 2026, and far more on certain query types: comparison queries trigger an AI Overview about 95% of the time, question-format queries about 86%, and "near me" queries about 77%. They are rarest on transactional queries, where Google protects the click path to a purchase.
Yes. A randomized field experiment from Carnegie Mellon and the Indian School of Business (published April 2026) found that removing AI Overviews increased outbound clicks by about 38%, and that zero-click searches rose from 54% to 72% when an AI Overview was present. An Ahrefs study of 300,000 keywords found click-through rate on the top organic result fell about 58% on keywords that trigger an AI Overview. Being cited inside the AI Overview partly offsets this: cited brands earn about 120% more clicks than uncited ones on the same results page.
No, not reliably. Moz found that 88% of Google AI Mode citations do not come from the organic top 10, and one analysis found 37% of AI-cited domains do not rank in traditional search at all. AI Overviews and AI Mode are separate retrieval systems that cite the same URL only about 14% of the time. Strong classic SEO is the foundation, but it no longer predicts whether an AI engine will name you.
Third-party validation. Muck Rack's analysis of 25 million links found about 84% of AI citations come from earned media: journalism, research, and reference sources, not a brand's own pages. Ahrefs found brand mentions across the web correlate about three times more strongly with AI Overview visibility than backlinks do, and that E-E-A-T signals predict citation far better than domain authority. The lever is being referenced across the web, plus clean structure and current content on your own site.
Not for consumer AI search, based on 2026 evidence. Crawl-log studies across hundreds of millions of AI-bot visits show the major engines skip llms.txt and crawl HTML directly, and Google stated in May 2026 that the file receives no special treatment. It still has real value for agentic and developer tools that fetch it, and it is harmless site hygiene, but it is not a citation signal for ChatGPT, Perplexity, or Google AI Overviews today. Treat anyone selling it as a citation lever with caution.
Three things. First, get the technical foundation right: fast site, clean schema, current content, and crawler access for the AI retrieval bots. Second, earn third-party mentions in the sources AI engines trust, because that is what 84% of citations come from. Third, fix your local entity data: only about 1.2% of businesses get recommended in ChatGPT versus 35.9% in Google's local 3-pack, and AI gets local business facts wrong about a third of the time, so consistent, accurate listings matter more than ever.
AI Search Optimization
Run your site through the free Meridian15 AEO audit, then see the full AI search optimization service. We build the foundation, earn the coverage, and track the citations.
Run the AEO Audit