Before launch, 99takes shows how personas from your audience respond to your copy or landing page: which version they prefer, what holds them back and how the page comes across. The jury is curated and runs on Claude; we don't have our own language model. Below is what you get back, in some places with a made-up example.
A buying committee reads copy differently from a family comparing two plans. Every audience falls into one of three types: business buyers, consumers or politics. The jury reads in whichever role fits.
For political copy, it judges whether a message lands. It doesn’t judge the position itself, and it doesn’t take sides.
Give 99takes the URL of your page, and it checks everything in one go: how the copy lands with your audience, how the structure and reading flow work, how the page comes across on desktop and mobile, and which of your core claims competitors make too. On top of that, a specialist reviewer looks at the page’s design craft.
You end up with an action plan, most important items first. If an action needs information only you have, such as a figure, Claude asks you for it instead of making something up.
This is the A/B test before launch. 99takes writes a challenger version of the opening or the whole page. You see it before the test and can edit it. Only then do the personas read your current page and the new version.
99takes tells you the price for each scope up front. It doesn’t replace an A/B test with real traffic; it shows which version is worth taking live.
Not every persona counts the same. You decide who your offer is made for (your core audience) and who is just reading along (your extended audience). The core audience decides the verdict; the extended audience is shown alongside it.
If a version wins only with the extended audience, you get a warning instead of a win. Otherwise you’d ship copy that convinces the wrong people.
Core audience: new version wins 4 of 4.
Extended audience: 1 of 2.
Some copy puts nobody off but is nobody's first choice. The scores look fine, yet the copy doesn't win anyone over. In that case the result carries a note, so you don't mistake bland copy for good copy.
Note: everyone likes it a little, nobody's first choice.
Add your brand promise in one sentence. Each persona says whether the copy keeps it or drifts away from it, with a short reason. This doesn’t affect the score. It’s a separate question: copy can win and still leave the brand behind.
Brand promise: 5 of 6 say "keeps it".
Add your page and copy from two or three competitors. For each of your core claims, 99takes shows whether a competitor promises the same thing (taken), something similar (contested) or nothing like it (open). Every rating comes with a verbatim quote and its source.
It also shows where an open claim sits on your page: in the hero or further down. A claim that three competitors also make won’t get you on anyone’s shortlist, however well it’s worded. This compares what the copy says, not how the market perceives the vendors.
Open: "fixed price per environment". Sits in section 8 of 11, belongs further up.
Taken: "publish without developers" (competitors A, B).
99takes also writes versions that compete with your current copy: from a run’s objections, from the customer voices and research behind your audience, or from scratch as a complete page based on your positioning and the story your copy should tell (old way, new way). On request, it uses a specific method, such as AIDA, benefit-led or headlines only.
99takes doesn’t invent customer voices or results. Any figure that doesn’t appear in the material is flagged for checking, and where proof is missing, the copy gets a placeholder. The versions are then tested like any other copy, and the jury doesn’t know where they came from. If a version was written from this same jury’s objections, the result says a win is expected and only the market can decide.
Version 2 · Proof first: "Fixed price per location, term [PROOF: minimum term]. No setup project."
Whether the copy convinces is one question. How the page comes across to your audience is another. In the design check, the personas see the whole page for themselves, on desktop and mobile. Each one says, from their own role, how the page comes across, what bothers them, what’s missing and what they’d change. Redesign suggestions are included. You get the screenshots, so you can see which page was checked.
This is the personas’ impression, not a measured result. It’s separate from the verdict on the copy, and unlike that verdict, it has no hit rate yet. If you want copy, structure and impression together, analyze the whole landing page.
Owner, trade business: "I only see the button for the first call after scrolling twice."
Two to four drafts of the first screen, as a URL, HTML or an image: light or dark, tiles or rows, with or without a photo. The personas say what registers with them, what they’d click and what makes them hesitate. The result shows which draft they’d click, or that it’s too close to call.
If you have five drafts, they don’t all need a full run. 99takes picks out the one worth running. It’s a shortlist, not a measurement, and the result says so.
A CFO reads copy differently when choosing a vendor than when justifying a renewal. So by default, every persona also reads the copy in the other personas’ situations. You see where the copy is weakest. If you want each persona only in its own situation, choose 1:1 and pay less.
Instead of testing a whole page at once, the build-out goes section by section, with one decision per section. The steps stay comparable with each other, even across several sessions. In the end, the finished page has to beat the old one.
Every run has a detail page with each persona's scores for both versions. A study becomes a report with its own link that you can send to colleagues or your client; in Slack or LinkedIn, the link shows a preview with the result. You can revoke a link at any time. On Pro and above, you can limit a report to invited readers.
Report which version won in the market. That builds up a hit rate per jury, always with the number of results and a range. It’s published only once there are ten decidable results; until then it reads "not measured yet".
The rate comes only from reported market results and belongs to the exact jury that made the call.
You can simulate a persona panel with any language model. What’s missing is anything that lets you rely on the result:
claude mcp add --transport http 99takes https://api.99takes.com/functions/v1/mcp