Audience research: personas built on evidence instead of guesswork.
99takes tests ad and website copy against your audience, right inside Claude. For the research, it reads your page and your competitors’ pages, finds public customer voices and builds your audience from them: 10 to 14 personas, each with its source. Everything 99takes writes and tests later builds on that.
Found
Draft: 12 personas
Four questions before the first run.
Read competitors
no creditsUp to five pages at once, including ones that only render in the browser: what they promise and what they cost. If you don’t know your competitors, 99takes suggests three to five for you to confirm.
“Who are our competitors, and what do they promise?”Find customer voices
no creditsPublic reviews, forums and discussions, each finding with its URL. These turn into pain points and objections. What vendors write about themselves doesn’t count as a customer voice.
“What do customers say about Example Supplier?”Set up your audience
15 credits10 to 14 personas with their jobs, clearly distinct even within the same role. Each one comes with its source; without evidence, a persona stays a hypothesis, and every answer says so. Refining individual personas is free. More on this: creating personas with AI.
“Build an audience for our page.”What only you say
3 to 7 creditsEach core claim is marked as taken, contested or free, with a verbatim quote and source. For contested claims, 99takes suggests a sharper angle that no competitor uses.
“Which of our claims do competitors make too?”Every persona says where it comes from.
Personas pulled from a language model’s memory sound plausible, but there’s no way to check them. With 99takes, every persona carries its evidence, and where there is none, it says so.
| What an audience brings | What you decide |
|---|---|
| Personas with role and job, 10 to 14, up to 24 | Who your core audience is and who is just reading along |
| Evidence for each persona: the quote and the URL where it was found | Which personas you refine; the rest stay unchanged |
| Pain points, positioning and narrative from the research | Whether you use a ready-made audience or set up your own |
| No evidence: marked as a hypothesis in every answer | What material of your own you add, such as objections your sales team hears |
From audience to first verdict.
Test copy against your audience
Personas from your audience read two versions or a whole page, and each gives a score and an objection.
Buyer simulation →See how your industry builds pages
B2B funnels broken down by how they’re built, plus a draft for your offer.
Funnel library →Your own audiences are private and visible only in your account. Without a plan, you can keep 1 audience of your own; the limits for each plan are on the pricing page.
Questions about audience research
What does an audience cost?
Setting up your own audience costs 15 credits once, when you save it. Questions, research and corrections up to that point are included, and refining individual personas afterwards is free too. How many of your own audiences an account can keep depends on the plan.
Where does the evidence come from?
From public sources: review sites, forums and discussions, each finding with its URL. On top of that comes anything you add yourself, such as objections your sales team hears. If nothing turns up for a persona, it stays marked as a hypothesis rather than getting made-up evidence.
How is this different from just having a language model write personas?
A language model will write plausible personas on request, but there’s nothing to check them against. With 99takes, every persona carries its evidence or a note that there is none. The audience stays in your account, so every run is judged by the same people.
Can I use a ready-made audience instead?
Yes. 99takes has ready-made audiences for common markets, such as mid-sized manufacturing, medical practices or home energy. Claude asks which one you want and never picks one for you. To test your own positioning, set up your own audience.