AI search meta-analysis Q4 2026: what 32 studies say

a close up of a computer screen with a blurry background

Every quarter, the evidence on AI search shifts enough that last quarter’s playbook needs revising. Kevin Indig of Growth Memo just published his Q4 2026 meta-analysis, drawing on 32 original studies covering consumer behavior, click data, brand citation mechanics, and agentic shopping. Here is what the data actually says, without the hype.

How much search has moved to AI?

The honest answer is: we can’t put a clean percentage on it. AI and traditional search have merged at the interface level, so separating “AI searches” from “Google searches” is increasingly artificial.

What the data does show:

  • A 2026 Pew survey of 5,119 U.S. adults found 49% had used an AI chatbot at least once, up 16 percentage points from 33% in 2024. 24% used one daily. 42% used chatbots specifically to search for information.
  • 60% of respondents reported reading an AI summary at the top of a search result.
  • AI Overviews are expanding into commercial queries. Semrush tracked 400 U.S. desktop commercial keywords and found Overview coverage rose from 45.75% in September 2025 to 88% in August 2026.
  • Profound classified 7.5 million U.S. English ChatGPT conversations and found commercial-intent conversations rose from 13.9% to 19.2% between June 2025 and June 2026.
  • Google still dominates. A Bain survey of 1,500 consumers found 56% used search engines as their default versus 16% for chatbots. SparkToro’s analysis of 41 search-enabled websites put Google at 73.7% of observed U.S. desktop searches in Q4 2025.

Indig’s verdict: consumers are adopting AI as a way to search, whether through AI Overviews or chatbots. No evidence exists for a cut in raw Google usage.

a person holding a cell phone in their hand

Do AI answers reduce clicks?

Yes, and the evidence is getting stronger. The key question is by how much, and the answer depends on the format and query context.

The landmark data point comes from a March 2025 Pew browsing panel of 900 U.S. users: users clicked a traditional result on 8% of visits with an AI summary present, versus 15% without one, roughly a 50% drop. They clicked a link inside the summary on only 1% of visits.

Two randomized field experiments add causal weight:

  • Agarwal and Sen’s 2026 field experiment found that when an AI Overview appeared, outbound organic clicks fell 39.8% and zero-click searches rose 34.5%. Sponsored clicks and overall search frequency remained unchanged.
  • Wang et al. ran a field experiment with 1,100 active U.S. participants and found that forcing searches into AI Mode reduced external click-through by 18.8 percentage points versus normal Google search.

Additional supporting data: Ahrefs analyzed 300,000 desktop keywords and found position-one CTR for keywords showing an AI Overview was 1.6% in December 2025, compared with a modeled 3.7% without one, a 58% relative gap. A difference-in-differences study of Wikipedia found a 5.45% reduction in English Wikipedia external-search referrals associated with AI Overviews.

One contrasting finding: Semrush and Datos examined more than 200,000 keywords and reported that for queries that began showing AI Overviews between May and October 2025, the measured zero-click rate fell from 33.75% to 31.53%. Indig notes that zero-click share and CTR are different metrics, so the studies are not necessarily in conflict.

What determines whether a brand appears in AI?

This is the most operationally useful section of the analysis. Research points to four distinct stages between a user query and a brand outcome, and brands can win or lose at each stage independently.

Stage 1: Discovery and retrieval

AirOps studied 15,000 prompts and 548,534 pages retrieved by ChatGPT. 85% of retrieved pages were never cited. Among cited pages that appeared in a top-20 search result, 32.9% appeared only because of a follow-up “fan-out” query, not the original prompt. That means tracking your brand’s visibility on the original query alone misses a significant portion of the retrieval process.

A Chen and Koudas study found that retrieval and topical relevance were more predictive of citations than features like FAQ blocks or schema markup. The SAGEO Arena benchmark added a nuance: reworking titles and meta descriptions improved retrieval rates, while rewriting body text to add technical terms or source references actually reduced them in benchmark tests.

Stage 2: Selection and citation

Getting retrieved doesn’t guarantee a citation. In Indig’s study with AirOps, the first retrieved page had a 58% citation rate, compared with 14% further down the retrieval list. Pages whose headings matched the query more closely also had higher citation rates.

A controlled study of 252,000 two-document trials found that relevance and source position had the largest effects on which source was cited first. SISTRIX found associations between citations and features like the current year in the title, a table of contents, and schema. Indig’s study with Gauge found that eight pages qualifying as primary research averaged 11.3 citations each, versus 3.4 for other pages.

One caution: in the C-SEO Bench, adding statistics lowered a document’s average ranking in 19 of 24 tested task-and-model combinations. Tactical page edits do not reliably produce citation gains.

Stage 3: Brand mention and framing

Being cited and being named are two separate outcomes. In Indig’s study with Semrush, 61.7% of 3,981 recorded domain appearances were citations without a brand mention, which the analysis terms “ghost citations.” Another 13.2% were both cited and mentioned, while 25.1% were mentioned without a citation.

Ahrefs’ analysis of 75,000 brands found a correlation between branded web mentions and AI visibility, though brand size and category could influence both independently.

Stage 4: User action and business outcome

In Indig’s Wynter survey of 200 software buyers, 72.5% reported visiting vendor websites after an AI chat. In Profound’s opt-in panel, brand-site visits within seven days were 2.07 percentage points higher after a ChatGPT mention than a forecast based on those same users’ prior behavior.

computer monitor

How much do buyers trust AI recommendations?

The data suggests AI both narrows and expands consideration sets, sometimes in the same buying session.

In Indig’s Growth Memo study with Citation Labs and Clickstream Solutions, 48 U.S. participants completed 185 recorded purchase-research tasks. The study found a 64% no-click rate in AI Mode and that the first-ranked item was the top choice in 74% of 137 codable AI Mode observations.

A Profound study had 56 people complete 224 ChatGPT shopping tasks. 92.8% ended without a meaningful open-web click. Selected brands had 24% share of voice in answers versus 11% for brands that were passed over. The reported correlation between visibility and choice was 0.57.

In the Wynter software buyer survey, 62.7% of respondents who already had vendors in mind added another vendor through AI. Among the 126 who reported choosing a vendor, 78 said it had appeared in the first answer.

That said, buyers do verify. In Attentive’s August 2026 survey of 3,054 U.S. adults who had recently browsed or shopped, 93% of AI-assisted shoppers said they sought outside information or proof while deciding whether to buy.

Are people ready to let AI shop for them?

Not really, and the technology isn’t quite there either.

Gartner’s January 2026 survey of 322 U.S. consumers found 31% were willing to let AI narrow household-supply choices and 28% were willing for personal electronics. Willingness to let AI make the actual purchase decision topped out at 11% across the lower-stakes categories studied.

On the capability side, Visa’s November 2026 shopping benchmark tested 22 AI models across 15 scenarios with three attempts each. Models passed 57.2% of attempts overall, and only 34.2% of model-and-scenario combinations passed all three attempts. The benchmark tested whether models could identify customer requirements and reach an appropriate buy-or-no-buy decision; it did not test checkout or payment.

Indig’s study with Siteline tested Claude Sonnet 4.6 agents across 1,500 runs for 100 B2B products. Pricing and features had a 79% first-party answer rate, compared with 93% for integrations and 92% for security. When no real price was disclosed, 45% of pricing runs cited at least one third-party source, versus 18% when a numeric public price was available.

What operators should do with this

Indig closes with three practical actions. These are his words, not ours, and they’re worth keeping:

  • Measure the journey as distinct stages: organic clicks, AI mentions and citations, qualified visits, leads, and sales are separate outcomes. Treat them that way.
  • Make decisions easier: give buyers and AI systems clear information about pricing, features, comparisons, and proof. Hidden pricing creates third-party fallback.
  • Test your visibility: run your important prompts repeatedly, check how your brand is described in AI answers, and monitor for changes in traffic and downstream business outcomes.

The full analysis covers 32 original studies published through November 13, 2026. The methodology section notes that studies use different populations, products, time periods, and measurement approaches, including experiments, surveys, browsing panels, modeled traffic data, and agent simulations. No single statistical estimate was computed across the body of work.

Stay on top of AI & Automation with BizStack Newsletter
BizStack  —  Entrepreneur’s Business Stack
Logo