Research · State of AI Search 2026

State of AI Search 2026: How 5 Engines Pick Their Citations, by the Numbers

A synthesis of public benchmark data + Resocial's audit corpus across 200+ brands. The citation share each major AI engine assigns to Wikipedia, Reddit, editorial sources, and direct brand content — plus where the asymmetries are widening.

Quick answer. Across 5 major AI engines (ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews), citation logic has bifurcated into two distinct strategies: editorial-source-weighted (ChatGPT, Claude, increasingly Gemini) and community-source-weighted (Perplexity, with secondary weight in AI Overviews). The single most-cited source across all engines remains Wikipedia (~32% weighted average), followed by Reddit (~21%), tier-1 editorial (~14%), and brand-owned content (~11%). The remaining ~22% is split across forums, YouTube, GitHub, primary research, and government/academic sources. Brand investment priorities should reflect which engines drive your specific audience — and the patterns reverse meaningfully between B2B and B2C buyer journeys.

Methodology

This report synthesizes:

  • Public citation benchmarks from Profound, Otterly, Athena HQ, Similarweb, and Conductor (2025-Q1 2026)
  • Resocial’s internal audit corpus of 200+ B2B and B2C brand citation samples across the 5 major AI engines (collected Q4 2025 through Q1 2026)
  • Manual sampling of 60 priority queries per engine, weighted across SaaS, fintech, ecommerce, healthcare, legal, travel, and B2B services verticals

Where Resocial’s internal data conflicts with public benchmarks, we’ve noted the divergence and explained the likely cause. This is synthesized research, not primary new data collection — the value is in the cross-source pattern recognition + senior analyst interpretation.

Headline numbers

Citation share by source type (weighted average across 5 engines)

Source typeWeighted citation share
Wikipedia32%
Reddit21%
Tier-1 editorial (Forbes, NYT, WSJ, BBC, trade pubs)14%
Brand-owned content (your site)11%
Government/academic (.gov, .edu, primary research)8%
YouTube transcripts6%
Other forums (HackerNews, StackOverflow, Quora)4%
GitHub / technical docs4%

The 11% brand-owned share is the part you can directly influence with on-site work. The remaining 89% requires off-site investment — Wikipedia/Wikidata entity, Reddit presence, tier-1 PR.

Citation share by AI engine

Where each engine pulls its citations from:

SourceChatGPTPerplexityGeminiClaudeAI Overviews
Wikipedia47.9%12.4%38.1%41.2%25.3%
Reddit6.1%46.7%8.3%4.5%18.4%
Tier-1 editorial18.2%8.1%14.6%21.7%11.9%
Brand-owned12.4%13.2%10.1%8.8%14.7%
Gov/academic6.3%7.1%12.4%13.8%5.2%
YouTube1.2%2.4%11.2%0.8%7.6%
Other forums5.1%6.3%2.8%6.4%4.8%
GitHub/docs2.8%3.8%2.5%2.8%12.1%

(Composite of public benchmarks + Resocial sampling. Sources don’t sum to exactly 100% due to multi-source overlap.)

The asymmetries

  • ChatGPT and Perplexity are inverse on Wikipedia vs Reddit (47.9% vs 6.1% on Wikipedia; 6.1% vs 46.7% on Reddit). A brand visible in one is often invisible in the other. We unpack this further in ChatGPT vs Perplexity for SEO.
  • Gemini is YouTube-heavy at 11.2% — far above any other engine. Video transcript optimization disproportionately rewards Gemini.
  • AI Overviews lean technical with 12.1% from GitHub/docs and meaningful gov/academic share. Reflects Google’s broader index distribution.
  • Claude is the most conservative — highest Wikipedia + tier-1 editorial weighting, lowest Reddit/Quora exposure. Closer to ChatGPT’s posture than Perplexity’s.

What this means by buyer journey

B2B research-heavy journeys (SaaS, fintech, enterprise services)

Journey stageDominant engine in 2026Optimization priority
Initial educationChatGPT, Perplexity, GeminiWikipedia entity + Quick Answer Block + comparison content
Vendor shortlistingPerplexity, AI OverviewsReddit presence + tier-1 editorial + comparison posts
Deep evaluationChatGPT, ClaudeCase studies + original research + named authors with credentials
Final decisionAI Overviews, branded searchKnowledge Panel + tier-1 press + LinkedIn presence

Implication for B2B: Wikipedia + Reddit + editorial PR are the three foundational investments. Brand-owned content is necessary but not sufficient.

B2C consideration journeys (consumer retail, travel, financial products)

Journey stageDominant engineOptimization priority
DiscoveryAI Overviews, GeminiSchema markup + product pages + reviews integration
ComparisonPerplexity, AI OverviewsSide-by-side comparison content + structured data
Purchase intentAI Overviews, brandedKnowledge Panel + Google Business Profile + trust signals

Implication for B2C: AI Overviews dominate; investment skews toward Google’s surfaces + structured data + reviews.

The freshness penalty (and what it means)

Public benchmark data confirms what we see in audits:

  • Content older than 12 months: 38% lower citation rate in Perplexity (controlling for topic + ranking position)
  • Content older than 6 months: 14% lower citation rate in AI Overviews
  • Content older than 24 months: 52% lower citation rate in Perplexity, 21% in AI Overviews
  • ChatGPT and Claude: freshness penalty is minor (under 10% for content up to 36 months old) when content has been minor-updated

The implication: a quarterly content refresh program on the top 20 pages is one of the highest-ROI interventions in AI Search Optimization. Many brands ship content once and never revisit; the citation share decays over 12-18 months.

Brand-owned content: what works

Even within the 11% brand-owned share, certain content patterns dominate. From Resocial’s audit corpus:

PatternCitation likelihood (vs baseline)
Quick Answer Block at top of page+4.2×
FAQ schema with structured Q-A pairs+3.1×
Comparison tables (side-by-side, 2+ entities)+2.8×
Source-cited statistics (e.g., “(HubSpot 2025)”)+2.4×
Numbered/bulleted lists+1.9×
Original data charts+3.5× (when content is fresh)
Named author with credentials+1.7×
Long-form (1500+ words) with cluster linking+2.1×

(Baseline = same page without the pattern. Sample = 1,200 cited pages across 200 brands.)

The patterns are mostly identical to what wins traditional SEO + AEO. The compounding effect is what’s new: a page with all of these patterns is cited at 6-12× the rate of a page with none.

We’ve broken down each pattern in The Complete Guide to AI Search Optimization.

1. Citation share is consolidating

Top 15 domains now capture 68% of all AI citations (up from 52% in early 2024). The first-mover advantage in entity authority is widening. New brands face a steeper climb than they did 18 months ago.

2. Wikipedia matters more, not less

Despite predictions of “post-Wikipedia AI,” every major engine continues to weight Wikipedia heavily. ChatGPT’s reliance grew slightly in late 2025 (from 44% to 47.9%). Wikipedia-eligible brands that haven’t pursued a page are leaving the largest single citation lever unpulled.

3. Reddit is becoming a tier-1 channel

Perplexity’s Reddit weighting drove a broader shift — Gemini and AI Overviews increased Reddit share by 4-6 percentage points in 2025. Brands without Reddit presence are increasingly invisible across multiple engines simultaneously.

4. Freshness penalties are sharpening

Engines are weighting fresh content more aggressively. Quarterly refresh cycles on priority pages are becoming standard practice.

5. Citation tracking tooling is consolidating

Profound, Otterly, and Athena HQ have emerged as the credible 2026 options. Major SEO platforms (SEMrush, Ahrefs) are adding capabilities but the data depth is shallow. Expect 2-3 acquisitions in this category before EOY 2026.

6. Engine-specific behavior is diverging

Some metrics show ChatGPT, Claude, and Perplexity drifting further apart in their citation logic over 2025, not converging. The “optimize for AI search” abstraction is becoming less useful; engine-specific tracks are becoming more necessary.

What to do with this data

For most brands in 2026:

  1. Identify which engines matter most for your audience (B2B research-heavy → all 5 weighted; B2C consideration → AI Overviews + Gemini; technical audience → Gemini + Claude + GitHub-adjacent)
  2. Map your current citation share across those engines (manual sampling of 30 priority queries)
  3. Identify the largest gap between current and benchmark — usually Wikipedia or Reddit for B2B brands, AI Overview structure for B2C
  4. Invest in foundational entity authority before engine-specific tactics
  5. Set up weekly citation tracking to measure progress

For an opinionated 90-day sequence applied to your specific situation, the free SEO audit returns an AI Search visibility baseline + prioritized action plan within 48 hours.

Caveats and limitations

  • Citation data is noisy. Different sampling methodologies produce 5-10% variance in headline numbers. Trust direction-of-change more than absolute percentages.
  • Industry effects are real. B2B SaaS citation distributions differ materially from B2C retail. Apply industry-relevant benchmarks rather than the cross-vertical average.
  • AI engines change behavior continuously. Numbers in this report reflect late Q4 2025 through Q1 2026 sampling. Major model updates (ChatGPT, Claude, Gemini) periodically shift weighting by 5-15%.
  • This is synthesis, not primary research. Underlying sources (Profound, Otterly, etc.) collected the raw data; Resocial added the analyst interpretation and cross-source aggregation. We plan to publish primary original research later in 2026.

Sources and further reading

External benchmarks consulted for this synthesis: Profound (citation tracker data), Otterly (citation tracker data), Athena HQ (citation tracker data), Similarweb (AI traffic measurement), Conductor (SERP feature tracking), HubSpot (conversion data), Forrester (B2B buyer behavior surveys).

Resocial-internal posts that go deeper on specific aspects of this data:

How Resocial uses this data

Every Resocial AI Search engagement starts with a citation baseline scan across the 5 engines. We compare client baseline to category benchmarks (the numbers in this report) and identify the largest gaps. The diagnosis informs which of the AI Search & GEO services to prioritize. We re-baseline quarterly to track progress.

For brands curious about their current citation share: a free SEO audit includes the AI search visibility scan at no cost.

Want strategy like this for your brand?

Get a free SEO audit

60+ dimensions, 48-hour turnaround.

Get a Free SEO Audit

Submit an enterprise RFP

Tailored proposal in 5 business days.

Submit an Enterprise RFP