5W ranks itself No. 1 in ChatGPT. The new “best of” list is self-written

Remember the “Top 10 agencies” lists that just happened to be published by one of the ten agencies? They are back. With prompts instead of a jury.

What 5W measured

US firm 5W Public Relations published a self-run analysis on November 1 and pushed it out via press release on November 6. The key numbers, per 5W:

• 396 unbranded prompts, questions a buyer might ask an AI, run roughly 140,000 times in October 2026.
• Six platforms: ChatGPT, Google AI Mode, Google AI Overviews, Gemini, Microsoft Copilot and Perplexity, measured with Profound's tools.
• Result: 5W appears in 30.3% of answers that name any PR firm at all. Edelman gets 21.2%, Weber Shandwick 10.6%. On Copilot, Edelman edges ahead.

Now the important part: 5W itself says the prompts were written around its own practice areas, so its standing is an upper bound, not an industry ranking.

Whoever writes the questions wins the ranking

That, to me, is the real insight. AI visibility has no neutral index everyone reads the same way. There is a list of questions. And the list decides the result.

You can see it in 5W's own data. In healthcare PR, another agency leads with 52.1% and 5W ranks third. In corporate communications and awards, Edelman is ahead. Different topics in the prompt set, different winner.

So when a vendor shows you a GEO study where they come out on top, ask for the prompts first. Not the percentage.

Visibility is not share of voice

One more detail that gets lost in the headline. The 30.3% is visibility: in how many answers the name shows up at all. 5W's share of voice, its slice of all brand mentions, is 4.9% in the same analysis.

Share of voice has been a planning staple since Binet and Field's work for the IPA. Treat AI visibility like classic share of voice and you are comparing apples to oranges. Both numbers are useful. They just answer different questions.

The catch: self-measurement becomes a PR weapon

5W disclosed its methodology, with limitations, time window and definitions. That is more than many do. Still, the press release carries the 30.3%, not the footnotes. I expect rankings like this to show up in every category in 2027. Software, consulting, agencies. And some of them will drop the footnotes.

Disclosure: I offer an AI visibility check on i4u.business myself. Which is exactly why I think a measurement is only as honest as the prompt set behind it.

What I would do now

• Build your own prompt set. 20 to 30 questions your customers actually ask, no brand names. Not the ones where you want to look good.
• Measure monthly, keep it fixed. Same prompts, same platforms. That is the only way to see movement instead of noise.
• Audit other people's rankings. Who measured, with which questions, over what period? Without those three answers, the number is an ad.

Bottom line: In AI search, it is not just who gets named that wins, but who writes the questions for the ranking.