Is Wikipedia a major source for LLMs?

0.39% of everything AI engines cite — about 1 citation in 258

Is Wikipedia a major source for LLMs?
Is Wikipedia a major source for LLMs?
RESEARCH

How much of what AI engines cite is Wikipedia? This analysis measures Wikipedia's share of citations across OpenAI, Perplexity, Gemini, Google AI Mode, and Google AI Overview, and breaks it down to find where Wikipedia actually functions as the canonical reference for an entity, and where it drops out entirely.

977 accounts5 providers7.6B citations
METHOD

The sample covers accounts measured across all five providers, so every figure is directly comparable. Each row is one citation.

Baseline

ACCOUNTS977
WIKIPEDIA SHARE0.39%
TOTAL CITATIONS7.6B
WIKIPEDIA CITATIONS29.6M
FINDING

Wikipedia is not a dominant citation source in AI answers. At 0.388% of all citations, its role is marginal in volume terms. Whatever anchoring function it serves is concentrated in specific query types rather than spread across the corpus.

The share is falling

0.388% ALL-TIME BASELINE
0.0%0.4%0.8%1.2%1.6%1.42%0.86%1.25%1.46%0.85%0.38%0.97%0.74%1.11%0.65%0.52%0.27%0.32%0.27%0.22%0.20%
MAYJUNJULAUGSEPOCTNOVDECJANFEBMARAPRMAYJUNJULAUG20252026
Wikipedia citation share by month. 977 accounts, five providers, May 2025 to August 2026. August is a partial month, shown hollow. Hover a month to read its value.

Wikipedia's share of AI citations fell from 1.46% to 0.22% over the past 12 months, roughly a sixfold decline. The drop is concentrated in a single step between March and April 2026, after which the share settles into a narrow 0.22–0.32% band and does not recover. Wikipedia is now cited at well under half the rate implied by the 0.388% all-time baseline.

Monthly values are read from the source analysis. The April 2026 step is the fall from 0.52% to 0.27%.

By provider

The same 977 accounts appear in all five providers, so differences below reflect engine behaviour rather than which companies were measured.

OpenAI0.96%1016M1.7B
Perplexity0.39%46.5M1.7B
Google AI Overview0.26%31.6M614.4M
Gemini0.23%2964K411.9M
Google AI Mode0.14%14.5M3.3B
Hairline marks the 0.39% all-provider baseline.

OpenAI used to be the outlier, and no longer is.

It opened the window citing Wikipedia at 5.32%, around 35× Google AI Mode and 50× Perplexity, and closed it at 0.29%, below Perplexity. The other four engines held flat in a 0.1–0.5% band throughout. Wikipedia's decline across the corpus is OpenAI's decline.

OPENAI, FIRST MONTH TO LAST5.32%0.29%
FURTHER FINDINGS
02

Even the highest provider is under 1%. Wikipedia is a minor citation source everywhere. Any anchoring role it plays is concentrated in specific query types, not spread across engines.

03

The other four LLMs show no trend. Perplexity, Google AI Overview, Gemini, and Google AI Mode each moved within a 0.1–0.5% band across the full 16 months and ended roughly where they started. They fluctuate, but in no consistent direction.

By industry

Healthcare0.51%4.8M929.1M
Travel & hospitality0.48%2M410.6M
All industries (baseline)0.39%29.6M7.6B
Financial services0.35%3.6M1B
B2B software & SaaS0.24%2.7M1.1B
Insurance0.17%349K202.2M
Hairline marks the 0.39% all-industry baseline.
FINDING

Travel & hospitality sits above the baseline, 24% higher, and the second-highest sector so far after healthcare. Airlines, major hotel chains, and well-known booking platforms (easyJet, Air France, Royal Caribbean, TripAdvisor, Agoda) carry substantial independent Wikipedia coverage, which likely drives the elevated share.

By prompt type

Same 977-account population, broken down by prompt type.

General0.45%5.6M1.3B
Informational0.42%8.6M2.1B
Commercial0.36%15.5M4.4B
Hairline marks the 0.39% all-provider baseline.
FINDING

Commercial prompts cite Wikipedia less than informational and general ones. Over the 15 months commercial sits at 0.366% against 0.450% for general and 0.421% for informational, and by July 2026 the gap is wider: 0.14% against roughly 0.35% for both. Commercial queries tend to pull vendor pages, comparisons, and reviews, where an encyclopedia entry is less useful as a source.

Figures are reproduced from the source analysis. Where the analysis quotes a monthly average that differs from the all-time table value, both are shown as written. Tables round shares to two decimals and shorten counts; charts and quoted findings keep the source precision. Hover a count for the exact figure.