The State of AI Search: Q2 2026 Data Study

Q2 2026 marks the inflection where AI-assistant search volume crossed 40% of B2B research starts. We tracked drift in citation patterns, ranking signals, and brand mentions across 5 AI platforms. The shifts since Q1 are larger than expected.

What We Did

We re-ran the 500-query panel from our Q1 study, adding 200 new queries in emerging verticals (AI tooling, dev tools, climate tech). Same 5 AI platforms. Same scoring rubric. Same 3-rater panel. We compared Q1 vs Q2 to surface drift.

Top-Line Findings

Finding 1: AI assistant share of B2B research starts reached 41% in Q2 2026.

Up from 33% in Q1. The fastest acceleration was in software evaluation queries, where 56% of evaluators now start with an AI assistant rather than a search engine.

Finding 2: ChatGPT now cites Reddit in 23% of consumer queries.

Up from 11% in Q1. Reddit emerged as one of the top 5 cited domains across consumer commercial queries. For B2B queries, Reddit citation is still under 5%.

Finding 3: Perplexity reduced citation count per response by 18%.

Average citations per Perplexity response dropped from 9.1 in Q1 to 7.5 in Q2. The model now favors fewer, more authoritative sources. Position #1 citation matters more than ever.

Finding 4: AI Overviews now appear on 41% of commercial intent queries.

Up from 28% in Q1. Click-through rate on position 1 dropped from 19% to 13%. Traditional SEO is producing fewer clicks even at the same rank.

Finding 5: llms.txt adoption tripled to 14% of top-100 sites.

Up from 4% in Q1. We measured no causal effect on citation lift yet, but adoption is now mainstream enough to be the baseline expectation.

What This Means for Your GEO Program

Treat AI assistants as a primary research channel, not a side bet. Reallocate at least 20% of search marketing budget to GEO if you have not already.
Audit your Reddit footprint if you sell to consumers. Authentic Reddit presence is now a meaningful citation lever for B2C.
Optimize for position 1 on Perplexity. With fewer citations per response, lower positions are dropping off entirely.
Accept declining CTR on traditional Google rankings. AI Overviews are siphoning clicks. The answer is to be cited inside the Overview, not to rank below it.
Deploy llms.txt. Table stakes by mid-2026.

Methodology in Detail

We sampled queries from three sources: keyword tools (Ahrefs, Semrush), reverse-engineered query logs from public AI search transcripts, and our own client panel of 40+ B2B brands. Each query was issued to ChatGPT (GPT-4o, 4.5), Perplexity (Sonar Pro), Gemini (2.5 Pro), Claude (Opus 4.5), and DeepSeek over a 6-week sampling window. Citations were extracted and normalized to canonical domain. Sentiment was scored by a 3-rater panel using a rubric (positive, neutral, qualified, negative). Inter-rater agreement was 0.81 on Cohen's kappa.

Limitations

This study reflects a single sampling window. AI models update retrieval and ranking continuously, so any specific finding may shift inside a quarter. We re-run the panel each quarter to track drift.

Want the Raw Data?

The query set, the citation logs, and the scoring rubric are available on request to vetted research and product teams. Request the dataset.

Frequently Asked Questions

How often do you update this study?

Quarterly. The next refresh will sample the same query set and report drift in citation share and platform behavior.

Can I cite this report?

Yes, with attribution to AY Rank and a link back to this page. We encourage citation in research, internal decks, and press.

Did you control for personalization?

Partly. Sessions were issued from clean profiles with no chat history. ChatGPT and Gemini still show some session drift even within clean sessions; we ran each query 3 times and report the modal citation.

Why these platforms and not others?

These five represent over 92% of AI assistant search volume in our internal panel. We omit niche platforms with under 1% share to keep the panel reproducible.

Ready to Get Cited by AI?

AY Rank builds the entity infrastructure, structured data, and citation-ready content that gets brands recommended by ChatGPT, Perplexity, and Gemini. Book a free GEO audit or see our case studies.

The State of AI Search: Q2 2026 Data Study