You'd judge anyone who cited Wikipedia for serious work. ChatGPT cites it more than anything else.
Ask ChatGPT to recommend a law firm and it leans hardest on the one source you would never put in a board deck or a due-diligence memo: Wikipedia. Not by accident. By design. Here is why, and what it means for whether your firm ever gets named.
Think back to the last time someone told you, out loud and with a straight face, that their source was Wikipedia.
Not for settling a bar argument. For something that actually mattered. A college thesis. A board deck. A due-diligence memo where one wrong fact detonates a deal and every citation had better be bulletproof.
What went through your head? Be honest. Somewhere between a raised eyebrow and "you're kidding, right?" Somewhere out there, a law professor just shuddered and has no idea why.
Now brace yourself: that is exactly what the AI your clients trust is doing. Right now, as you read this. Ask ChatGPT to recommend a law firm and it does not sheepishly sneak a glance at Wikipedia and hope nobody noticed. It strolls right up and leans on it harder than on anything else on the internet.
The data is blunt about it
Across the largest studies of AI citations run in 2025 and 2026, one source sits at the top of ChatGPT's list. It isn't a newspaper. It isn't a legal directory. It's Wikipedia.
Profound analyzed 680 million AI citations between August 2024 and June 2025. Wikipedia came out as ChatGPT's single most-cited source: 7.8% of everything it cites, and nearly half (47.9%) of its top-10 sources. Ahrefs, looking across 78.6 million searches, found Wikipedia the most-mentioned domain in all three major AI assistants: ChatGPT, Perplexity, and Google's AI Overviews.
And the prestige outlets you'd actually trust on something important? In a Q1 2026 analysis of roughly 600,000 ChatGPT citations, the Wall Street Journal, The New York Times, and Bloomberg did not crack the top 20 at all. A crowdsourced encyclopedia did. By a mile.
This isn't a bug they're racing to fix
It is by design. In GPT-3, the last major model whose training recipe was made public, Wikipedia was fed to the model 3.4 times over, more than any other source in the entire mix, while the massive open-web crawl was sampled less than once. The Wikimedia Foundation said it plainly in October 2025: almost all large language models train on Wikipedia datasets.
Why would the smartest systems on earth lean this hard on Wikipedia? Not prestige. Verifiability. Wikipedia is structured, neutral in tone, sourced line by line, updated constantly, and corroborated across thousands of other pages. That is the profile an AI trusts. Not because it looks impressive, but because it is checkable.
Hold onto that word. It is the whole game for your firm.
Why a managing partner should care
Legal is the most AI-exposed corner of search there is. 78% of legal search queries now trigger a Google AI Overview, the highest rate of any industry, per Semrush's analysis of more than 10 million keywords. Translation: your prospects are getting an AI answer before they ever reach a website. And more than half of consumers say they have used or would use AI for a legal question; of those who did, 28% were pointed toward contacting a lawyer (Clio, 2025). The machine isn't just answering. It is routing clients.
So who does it name? Mostly, not the firms themselves. When 5W and the Haute Lawyer Network tested how ChatGPT, Claude, Perplexity, and Google AI Mode recommend lawyers, the answers came from roughly seven directories: Chambers, Legal 500, Super Lawyers, Best Lawyers, Martindale, Avvo, Justia. Ask the AI about Cravath M&A expertise and the top results were Chambers and Legal 500 profiles. Third-party directories, outranking one of the most prestigious firms in America for its own practice area.
The firms inside the shortlist will compound. The firms outside it will not.
Ronn Torossian, 5W
"So we just make a Wikipedia page?"
No, and please don't try. Wikipedia rejects self-serving firm entries, and gaming it isn't the move. The move is to become the kind of entity AI can verify, everywhere it looks.
The data backs this hard. Ahrefs studied 75,000 brands and found the single strongest predictor of showing up in AI answers wasn't backlinks, a weak 0.22 correlation. It was how often the brand is mentioned across the web, a 0.66 correlation, three times stronger. Brands in the top quarter of web mentions earned roughly 10 times more AI mentions than the quarter below them. The bottom half were, in Ahrefs' word, essentially invisible.
Harvard Business Review has a name for the new scoreboard, share of model, and frames the rule without flinching: an AI's view of your firm depends on the independent evidence they can find about you, not on what your website says about itself.
That is the line to tape to the wall. Your website says you are the answer. So does everyone's. Wikipedia, the directories, the bylines, the third-party coverage: those say it about you, consistently, in a form the machine can confirm. The difference between marketing and evidence is the difference between being seen and being cited. AI weighs evidence.
The difference between marketing and evidence is the difference between being seen and being cited. Your website says you are the answer. So does everyone's. AI weighs the evidence it can verify.
Jacob Shamis, Co-Founder, Selectio.ai
The uncomfortable math
Three-quarters of legal searches now return an AI answer. Your buyers are already asking. The only open question is whether the machine has enough verifiable, consistent evidence about your firm to say your name, or whether it hands your next client to a directory and a competitor instead.
Most firms are still pouring budget into the website the AI barely reads, and ignoring the footprint it actually trusts. That gap is the whole opportunity right now, and almost nobody in legal has closed it yet. The window is open. It won't stay that way.
Sources: Profound, AI Platform Citation Patterns (680M citations, Aug 2024 to Jun 2025); Ahrefs, Most-Cited Domains Across AI Assistants (78.6M searches) and AI Overview Brand Visibility Factors (75K brands); 5W Citation Source Audit, Q1 2026 (~600K ChatGPT citations); Brown et al., GPT-3, "Language Models are Few-Shot Learners"; Wikimedia Foundation, New User Trends on Wikipedia (Oct 2025); Semrush, analysis of 10M+ keywords (legal AI Overview trigger rate); Clio 2025 Legal Trends Report; 5W & Haute Lawyer Network, 2026 Legal AI Visibility Report; Harvard Business Review, Preparing Your Brand for Agentic AI.
Frequently asked questions
Wikipedia. In Profound's analysis of 680 million AI citations between August 2024 and June 2025, Wikipedia was ChatGPT's single most-cited source: 7.8% of everything it cites and nearly half of its top-10 sources. Ahrefs, across 78.6 million searches, found Wikipedia the most-mentioned domain across ChatGPT, Perplexity, and Google's AI Overviews. Prestige outlets like the Wall Street Journal, The New York Times, and Bloomberg did not crack ChatGPT's top 20 in a Q1 2026 analysis of roughly 600,000 citations.
Not prestige, verifiability. Wikipedia is structured, neutral in tone, sourced line by line, updated constantly, and corroborated across thousands of other pages. That is the profile an AI trusts, because it is checkable. It is also by design: in GPT-3, the last major model whose training recipe was public, Wikipedia was fed to the model 3.4 times over, more than any other source.
No. Wikipedia rejects self-serving firm entries, and gaming it is not the move. The goal is to become the kind of entity AI can verify everywhere it looks: consistent, corroborated mentions across directories, bylines, and third-party coverage. Ahrefs studied 75,000 brands and found the strongest predictor of appearing in AI answers was how often a brand is mentioned across the web, three times stronger than backlinks.
Independent evidence, not your own marketing. Ahrefs found web mentions correlated with AI visibility at 0.66, versus a weak 0.22 for backlinks; brands in the top quarter of web mentions earned roughly 10 times more AI mentions than the quarter below. Harvard Business Review calls the metric share of model: an AI's view of your firm depends on the independent evidence it can find about you, not on what your website says about itself.
Legal is the most AI-exposed category there is. According to Semrush's analysis of more than 10 million keywords, 78% of legal search queries now trigger a Google AI Overview, the highest rate of any industry. More than half of consumers say they have used or would use AI for a legal question, and of those who did, 28% were pointed toward contacting a lawyer (Clio, 2025). The machine is not just answering; it is routing clients.
Which sources is AI reading about your firm?
Your free, live AI visibility check shows you exactly which sources AI is reading, where you stand across ChatGPT, Perplexity, Claude, Gemini, and Copilot, and which firms it is quietly handing your next client to instead.
Run your AI visibility check