How did we measure helloAISearch's own AI visibility, and what did we find?

1. 1Prompts: 7 category questions, written the way a buyer types them, frozen for the period.
2. 2Engines: ChatGPT, Claude, and Gemini. Perplexity was not in the account's plan for the period, so it is not in the count.
3. 3Cadence: the plan's schedule, weekly plus daily slots for the highest-priority prompts.
4. 4Counting: an answer counts as naming a tool if the tool's name appears in the answer text. Position and sentiment were recorded but are not in the headline number.
5. 5Share of voice: mentions of a tool divided by mentions of all tools tracked, over the period.

Those are the only competitor numbers on this site, and they are counts of names in answers, not claims about what those tools do. Their descriptions elsewhere on the site are limited to how the engines characterised them in the same answers.

• The engines recommended tools using the same five criteria in almost every answer: queries the engines directly, shows the answer text, measures share of voice, says what to do next, and is affordable for a small team. We meet them; we had not said so in a way an engine could quote.
• The winning tools had pages whose titles were the prompts. We had a home page.
• Consistency was our weakest signal. Our category was described three different ways across our own pages.

So the moves were the obvious ones: one direct-answer page per losing prompt, a glossary that defines our category in quotable sentences, an llms.txt, and the same Organization and Brand nodes on every page. This article is one of those moves.

The same count for the next period, including Perplexity, whatever it says. If the number moves, the pages worked. If it does not, we will say that too. Until a new period is closed, the scorecard above is the only number we quote, and it is dated.