Month two: what changed in the AI answers

A month ago I logged what ChatGPT, Claude, Perplexity and Gemini said about They Said. (Read month one →.) The point of doing this in public was always going to be the second month, not the first.

Four weeks have passed. I re-ran the same twelve prompts. Here's what shifted.

What changed: entity prompts

Last month, ChatGPT and Claude both returned a clean canonical description; Perplexity added editorial; Gemini was thin.

This month:

• ChatGPT added a one-line description of the four outcome services unprompted - coverage, AI visibility, awards, growth. That language is straight off the homepage. The Organisation JSON-LD is being read.
• Claude added the founder credit ("Fayola Douglas, founder of They Said") inside the synthesised answer when asked who runs the company. The author chip on the journal posts is feeding through.
• Perplexity cited the They Said website plus one third-party citation - an interview that ran on a podcast feed earlier this month. First external authority signal.
• Gemini is still thin. Gemini weights Google's open index more heavily and the site is still new in that index. This is the curve to watch.

The interesting result is that the consistency move (writing the same canonical entity description in the same words everywhere - site, schema, social bios) is being rewarded. The engines that synthesise are repeating our language back. That's not a coincidence - that's the entity-hygiene mechanic working.

What changed: discovery prompts

Last month: zero mentions on discovery prompts.

This month: one mention. Claude included They Said in a list of "smaller, founder-led content agencies serving B2B service businesses" when asked for recommendations. ChatGPT did not. Perplexity did not. Gemini did not.

One mention out of twelve possible is not a moment to celebrate. It's a vector. The mention was reasoned - Claude explained why it included us, citing the published-in-public AI work and the focus on agency clients. That's the engine starting to understand the position.

What changed: authority prompts

Still zero mentions on category authority prompts. Expected. Authority builds over months and is mostly driven by third-party citations - the trade press coverage we don't have yet, in volume.

The first earned-media mention from month two helped Perplexity. It hasn't moved the others. This is consistent with how the engines work - third-party citations compound, but the compounding has to clear a threshold before it shows up in synthesis.

What we're doing differently in month three

Two changes:

One. Doubling the pace of cornerstone publishing on AI visibility specifically. The pages that exist are useful; there aren't enough of them yet. The engines have to read fifty pages of evidence to confidently cite a position. We're closer to ten.

Two. Targeted trade press push. The hypothesis after two months of data is that earned coverage moves Perplexity fastest, ChatGPT next, Claude after, and Gemini last. So the next four pitches will all land in trade press that's indexed quickly and quoted often (Campaign, The Drum, PRWeek).

What this tells you (if you're considering an AI-visibility programme)

A few honest things.

It's slower than people sell it as. Anyone offering month-one results is not running the programme they're describing. The engines do not refresh that fast on a new entity.

Entity hygiene is cheap and underrated. The biggest movers in month two were the canonical description fixes and the Person schema on the About page. Both took a morning. Both showed up in the synthesised answers. If you do nothing else, do that.

Citations compound on a threshold. One earned-media mention moves one engine. Five mentions move three. There's a critical mass somewhere between five and fifteen citations where the engines start mentioning you on discovery prompts unprompted. We're not there yet. We will be.

Public tracking is the differentiator. Anyone can claim AI visibility results. Almost nobody is showing the engine answers month-by-month. The clients we're talking to about this service have all referenced these posts in the call. The series isn't just measurement - it's the deliverable.

Back in four weeks for month three. Expecting Perplexity to be the leader by then. We'll see.