Message testing: personas from your audience tell you which version holds up
Message testing means showing a message to people from your audience before it goes live, to learn whether it is understood, whether it convinces and what is missing. There are four ways to do it, from panels to simulated audiences. 99takes is the fourth way: personas from your audience read your versions right inside Claude.
Last updated:
What is message testing?
Message testing is a test of the message itself: is it understood, is it believed, and what is missing? Typical subjects are headlines and sublines, value propositions, landing page sections, ads and subject lines. The readers should arrive the way your buyers do: with the same question, the same prior knowledge and the same doubts.
A good test answers four questions: Do people understand what is offered and for whom? What stays unclear? Which version would they choose, and why? Which objection keeps them from the next step?
The four ways compared
| Way | Who reads | What you get | Strength | Limit |
|---|---|---|---|---|
| Panel | Real people from a panel, e.g. verified B2B professionals | Open-ended answers and ratings on your page or message | Real voices, good for high-stakes decisions | Time and cost per test; the panel has to cover your audience |
| A/B test | Real visitors to your page or ad | Difference in clicks or conversions | Measures real behavior | Needs enough traffic and time; doesn’t tell you why |
| Interviews | A few real buyers | Reasons, doubts and the buyers’ own words | Depth, new insight | Only a few voices; takes effort to organize and analyze |
| Simulated audience | AI personas modelled on your audience | Objections, choice and reasoning per persona, comparison of several versions | Fast and repeatable; works before launch and without traffic | Not a market measurement; only as good as the audience behind it |
Panel providers include Wynter and NewtonX (both listed by Gartner Peer Insights under “B2B Message Testing Solutions”). In its overview, Perspective AI divides the field in a similar way: panel, preference test, in-product intercept and simulated audience.
When to use which
If you have several versions before launch, a simulated audience is a quick way to shortlist them. Then test the best one in the market.
With plenty of traffic and a question about conversions, an A/B test is the right tool. It measures what happens, but not why.
When a lot is at stake, such as a new positioning, a relaunch or a new category, get real voices from a panel or interviews. A simulated shortlist beforehand means only the strongest versions take up real people’s time.
If you don’t know your buyers’ language yet, start with interviews or public customer voices such as reviews and forum posts. Write your versions from what you find, then test them.
Good questions for a message test
- What is offered here, in one sentence?
- Who is this made for, and are you one of them?
- What is unclear or hard to believe?
- What keeps you from taking the next step?
- Which version would you choose, and why?
Message testing with 99takes
You work in Claude (claude.ai or Claude Code) and share the copy or your page’s URL. Personas from your audience read each version and say which one they choose and where they get stuck. 99takes builds the audience from your competitors’ pages and public customer voices: typically 10 to 14 personas, backed by evidence wherever it exists. There are also ready-made audiences.
If you want, 99takes also writes challenger versions to compete with your current copy. They contain no unsupported numbers and are tested blind. The verdict follows the personas your offer is made for; the core audience and the extended audience are reported separately. Comparing two versions with 12 personas costs about 73 credits, or about 19 as 1:1. You start with 350 credits, no payment details.
How often the jury was right comes from market results that users report: a hit rate per jury, always with the number of results and a range. We publish it once there are ten decidable results. Without an account you can try the Headline Analyzer.
What a simulation cannot do
- It does not replace market measurement. If you have traffic and conversions, your own numbers are the better source.
- It is only as good as the audience behind it. Personas without evidence are hypotheses, and 99takes labels them that way.
- When two texts are compared, only the copy counts. How a page comes across on desktop and mobile is a separate question, answered by the design check.
Frequently asked questions
How much does message testing cost?
It depends on the method. Panel providers publish their own prices. With 99takes, comparing two versions with 12 personas costs about 73 credits, or about 19 as 1:1. You start with 350 credits and no payment details; credits start at €17.85 with no subscription (incl. VAT).
Is AI message testing GDPR compliant?
Ad and page copy rarely contains personal data, and the personas at 99takes are AI characters, not real people. Your copy and results are stored in the 99takes database at Supabase in Frankfurt, Germany. The analysis runs on Claude by Anthropic as a processor, also outside the EU under the EU Standard Contractual Clauses, and according to its terms not for training. Every account can use the data processing agreement (German version binding). Keep customer names and addresses out of the copy you test. Details are in the privacy policy.
Do I need traffic for message testing?
For an A/B test, yes. Panels, interviews and a simulated audience need no traffic and work before launch.
Does a simulated audience replace real customers?
No. It shortlists versions and surfaces objections before real people or ad budget are spent. Only the market can show whether a version wins there.
How accurate is 99takes?
99takes measures accuracy against reported market results, as a hit rate per jury, with the number of results and a range. We publish it once there are ten decidable results; until then, 99takes doesn’t state one.
Can I test ads and emails too?
Yes. 99takes checks headlines, ad copy, subject lines, emails, landing page sections and whole pages by link.
Related reading
Test the next version before it goes live
Create an account, connect 99takes to Claude and have your first piece of copy checked.
Start for free