Jonathon's AI Wiki

Similarweb 2026 Generative AI Landscape — The Evolution of AI Search

11 min read

Source: raw/similarweb-2026-generative-ai-landscape-evolution-ai-search.pdf — Similarweb, “2026 Generative AI Landscape: The Evolution of AI Search,” 38-slide report by Laurie Naspe (Director, Market Insights) + Sam Sheridan (Data Analyst, Market Insights), data through May/June 2026. Guest experts: Harry Clarkson-Bennett (DJB Strategies), Michael Horrocks (Miro), Ethan Smith (Graphite), Kevin Indig (Growth Memo), Rand Fishkin (SparkToro), Lily Ray (Amsive), Aleyda Solis (Orainti), Harel Amir (Similarweb Ad Intelligence).

Similarweb’s flagship annual synthesis of the gen-AI market, in three movements: platform evolution (adoption, market share, embedded AI), user behavior inside gen-AI sessions (demographics, prompt behavior, distinct audiences), and “winning in AI search” (brand visibility, citations vs referrals, and the arrival of ChatGPT advertising). Scope caveat stated up front: web + app only — the report excludes API usage, desktop applications, and in-app AI features, so every number here undercounts total AI activity.

Key Takeaways

  • Gen AI is mainstream infrastructure, still compounding. 9.5B average monthly web visits worldwide (Jun 2025–May 2026, +70% YoY, from 5.6B); 655M average monthly unique visitors (+57%, from 418M); 4.4B app downloads (+58%, from 2.8B). AI chatbots (0.65B UVs) now sit alongside Search (3.3B, +2%), Social (2.4B, +7%), eCommerce (2.2B, +7%) and above-trend against News (1.4B, −5%) — the fastest-growing major web category by far.
  • ChatGPT still leads the web but fragmentation is real. ChatGPT’s share of gen-AI chatbot web visits fell from ~83% (Jun 2024) to ~53% (May 2026) as Gemini, Claude, Perplexity, and DeepSeek expanded. In US app MAUs (May 2026 vs June 2025): ChatGPT ~50M (+87%), Gemini ~30M (+31%), Meta AI +435%, Claude +349%, Grok +117%, Perplexity +94%, while Copilot 365 (−31%) and DeepSeek (−23%) lost ground.
  • The future of AI is embedded. AI is integrating into every major category (Google AI Mode / Bing Copilot in search; Meta AI / WhatsApp AI in social; Alexa for Shopping and Shopify Sidekick in commerce; Microsoft 365 Copilot / Google Workspace AI in productivity; Perplexity Comet / Opera AI in browsers). Meta AI’s publicly reported monthly users more than tripled in 18 months: 384M (11% of Meta-family DAUs, Sept 2024) → 1.0B (29%, Oct 2025) → 1.2B (33%, Mar 2026). Alexa for Shopping (formerly Rufus) PDP views ran ~7x their June-2025 baseline by December 2025 and ~5.9x in May 2026, already routing meaningful traffic to brands like CeraVe, Skechers, Samsung, and Lego.
  • Google is funneling search into AI-native experiences. AI Overviews appeared on ~16% of US searches in Jan 2025 and ~42% by May 2026 (volatile in between: ~41% Aug 2025 peak, ~27% Oct 2025 dip); AI Mode visits climbed from ~120M (Jun 2025) to a ~320M peak (Mar 2026), settling ~270M in May 2026.
  • AI complements rather than replaces search. 95% of ChatGPT’s 494M web UVs also use Google (461M overlap; Google is 3.3B) — unchanged since Sep 2025. Cross-assistant multi-homing is high (ChatGPT 494M / Gemini 310M / Claude 86M; 29M use all three, Mar–May 2026).
  • Half of gen-AI users are now 35+. US chatbot usage share by age (May 2024 → 2025 → 2026): 18-24 fell 28→24→22% and 25-34 fell 33→29→28%, while 45-54 rose 12→14→16% and 55+ rose 12→16→18%. The 18-34 share dropped from 61% (May 2024) to 50% (May 2026) — Michael Horrocks reads growth spreading into older generations as “what durable, mainstream adoption looks like.” Counter-signal: early Gen-Z AI fatigue (graduation-speech backlash; DuckDuckGo’s noai.duckduckgo.com skews heavily 18-24: 33% of its users vs 16% of the US baseline).
  • Platforms are differentiating into distinct roles and audiences. Prompt behavior (US, Jun 2025–May 2026): Claude leads both median prompt length (~17 words) and prompts per session (~5.5) — the deep-conversation end; ChatGPT combines depth with breadth (~13 words, ~5 prompts); Perplexity holds the research middle; AI Mode (~9 words, ~2.7) and Gemini / Copilot 365 (~2 prompts) serve quick answers and tasks. Audience affinity vs Google baseline: ChatGPT = everyday consumers (Restaurants 2.4x, Health 2.3x, Fashion 2.1x); Claude = professionals and students (Universities & Colleges 25x, Programming 3.0x, Business Services 3.0x, Jobs 2.7x); Gemini = tech enthusiasts (Graphics & Web Design 4.8x, Computer Security 3.1x).
  • AI prompting habits are reshaping Google queries. Average US Google search length was flat (~3.33–3.36 words) until AI Mode launched (May 2025), then climbed to ~3.51 words by Apr/May 2026; searches over 5 words grew from 13.6% to 16.1% of queries YoY (Ethan Smith: users are adopting “a new more natural way to search and discover”).
  • AI recommendations shape the shortlist — the 2-4x lift. Users who saw a brand recommended by AI were two to four times as likely to visit that brand as a non-recommended competitor (US desktop, Jul–Dec 2025; shown across finance AmEx/Capital One, travel Skyscanner/Kayak, beauty Sephora/Ulta). Rand Fishkin frames it as billboard-era lift measurement reborn: “influence is happening. What marketers must do now is change the way they run measurement and attribution.”
  • Citations went from rare to routine in under a year. The share of ChatGPT answers containing web citations rose more than fivefold, from ~1.2% (Jun 2025) to 6.8% (May 2026). It is intent-dependent: Travel 22.6%, Retail 13.5%, Sports 10.7%, Finance 8.0% vs Technology 6.6%, Health 6.3%, Education 4.8% — AI searches the live web most when users compare products or need current information.
  • The citation mix depends on the question — there is no universal GEO play. Top-10K US citations, May 2026: ChatGPT overall cites Reviews & UGC 28.9%, News & Publishers 26.0%, Retail & E-commerce 14.1%; AI Mode leans News 24.3% / Reviews 19.9% / Social 10.9%. By industry: Beauty conversations cite Retail & E-commerce 54.7%; Travel cites Reviews & UGC 54.1% + Consumer Services 20.5%; Finance cites specialist Finance sources 36.6% + News 28.0%. “Brands need to understand which content types AI trusts in their category.”
  • Links drive traffic; citations build trust — and they live at different depths. 65% of ChatGPT-cited URLs sit two or three folders deep (depth 2 alone = 41.7%), while 58.8% of referral traffic lands on homepages. Aleyda Solis: strengthen the deep pages that influence AI answers, and make the top-level pages that receive the clicks conversion-ready — they are different jobs, assess them separately.
  • ChatGPT’s May 7 (2026) search update reshaped referral behavior. Referral volume jumped ~90% above its March baseline within weeks, and the share of referrals landing on homepages more than doubled, from ~25% to ~60% — ChatGPT became “a gateway to brands themselves” rather than to individual content pages.
  • AI referrals are growing fast off a small base, and AI favors platforms. Worldwide (Jun 2025–May 2026): Marketplaces 46.8M referral visits (Amazon + Temu; +237.3% YoY), News 44.5M (+115.6%), Travel 44.5M (+115.6%), Finance 19.1M (+236.5%), Consumer Electronics 17.1M (+215.4%), Fashion 7.1M (+278.2%), Beauty 2.0M (+312.5%). Smaller-base verticals grow fastest; “AI favours large, multi-category platforms over smaller, vertical-specific sites.”
  • Brand visibility has early, category-specific leaders. Most Visible Brands in GenAI (US, May 2026, beauty tab): CeraVe 100 (indexed leader), Ulta 86, Sephora 80, La Roche-Posay 74, The Ordinary 67, then a steep drop (e.l.f. 33, Maybelline 23). Kevin Indig’s case for Share of Voice: “users focus on brand mentions instead of clicking on sources. Visibility is success.”
  • Lily Ray’s RAG argument: GEO rests on SEO. Every major AI search product relies on retrieval-augmented generation, so “if your organic visibility dips, your AI search visibility will follow.” Brands that accelerated AI-generated content production to chase AI-search visibility hurt their SEO and are now declining in both — “an even steeper climb back… than if they’d never chased the shortcut.”
  • A ChatGPT advertising ecosystem arrived within a single quarter. Ad penetration hit 26% of US ChatGPT chats by June 2026 (vs 14% in May); the share of conversations containing ads climbed through Apr–Jun 2026 across the US (~10%→~30%), Australia, Canada, the UK (~0 to ~35% within June), and Japan. Top US advertisers (June 2026): Resume.io, Monday.com, Framer, Axon by AppLovin, Attio, BestMoney, Aikido Security, SAP, Home Depot, VistaPrint. Mechanics: 66.3% of ads appear after the second prompt or later (context-targeted “Ad Moment”), 65.3% of sessions continue after an ad appears, CTR ~0.50%.
  • The closing thesis (Baruch Toledano): a new era of discoverability. Every AI interaction is a potential discovery moment; fundamentals (recognized brand, trust, authoritative content, omnichannel presence) matter more than ever, but measurement must connect visibility to downstream outcomes. “Visibility is no longer about ranking in one place; it’s about being discoverable wherever decisions are made.”

Where this sits in the wiki’s evidence stack

^[inferred — wiki-side positioning, not stated in the report]

This is the market-context / user-behavior layer of the AI-citation thesis cluster, refreshed through May/June 2026 and superseding several scattered data points with one coherent snapshot. It corroborates and extends the existing Similarweb cluster: the visits/referrals decoupling of AI Search Trends (visits 9.5B/month vs referrals in the tens of millions per category — still a rounding error, but now growing 2-3x YoY), the ~53% ChatGPT web share of the chatbot traffic data, the 2.5x downstream lift of Downstream Impact (now framed as 2-4x subsequent visits), and the brand-mention thesis of the January Brand Visibility Index (new May 2026 category leaderboards). Two findings are genuinely new to the cluster: the May 7 ChatGPT homepage-referral shift (25%→60% of referrals landing on homepages, while 65% of citations come from depth-2/3 pages — the sharpest evidence yet that citation strategy and landing-page strategy are different problems), and the quarter-speed arrival of conversational advertising (14%→26% penetration in one month).

Try It

  1. Split your AI-search reporting into the two jobs the folder-depth data separates: track cited deep pages (evidence) and homepage conversion readiness (traffic) as different metrics with different owners.
  2. Look up your category’s citation mix before writing a GEO plan. Beauty = e-commerce pages, travel = reviews/UGC, finance = specialist publishers; assume nothing transfers across verticals.
  3. Use the 2-4x recommendation lift and the 35+ demographic shift as the client-education numbers when a stakeholder still frames AI search as a young-user niche that doesn’t influence buying.
  4. Watch ChatGPT ad penetration quarterly (26% of US chats, June 2026) — at current trajectory, conversational ads become a plannable performance channel; the top-10 advertiser list is already SaaS-and-retail-heavy.
  5. Quote Lily Ray’s RAG argument to anyone proposing AI-generated content shortcuts for AI visibility: organic visibility is the retrieval substrate; hurting SEO to chase GEO burns both.
  • weomarketly:ops/similarweb-generative-ai-landscape-2026-07: the WEO-lens extract compiled 2026-07-31 from the same PDF (dental translations: the dead age objection, the Visibility Scan sales stat, reviews as citation infrastructure, homepage-first corroboration).

Open Questions

  • Health/dental citation mix is missing. The per-industry citation-mix breakdowns cover beauty, travel, and finance only; health appears in citation frequency (6.3% of answers) but not in mix. Which source types AI trusts for health/dental queries remains unmeasured here (the work vault runs this as an OpenLens test candidate).
  • Ad Moment measurement is one month old. The 26% penetration, 0.50% CTR, and 66.3% post-second-prompt figures are June-2026 US desktop snapshots of a channel that changed month over month; treat as direction, not baseline.
  • The web+app scope excludes API, desktop apps, and embedded in-app AI — precisely the surfaces the report itself argues are the future (Meta AI, Copilot, Alexa). Total AI influence on discovery is strictly larger than every figure above.
  • AI Mode visits dipped Apr–May 2026 (~320M → ~270M) with no explanation offered; watch whether that is seasonality, cannibalization by AI Overviews’ climb to 42%, or measurement noise.

Graph View

Backlinks