Freemium Planning & Strategy

Minds

Message and concept testing against AI-simulated audiences, priced so a small team can actually try it

Visit Minds Verified Sep 2026 · Next review scheduled Dec 2026

Minds builds reusable AI personas from your own uploads, links and public data, then runs interviews, group questioning and structured questionnaires against them, returning both aggregate distributions and individual responses. Message, concept and ad pretesting is a first-class function rather than a repurposed user-research feature, covering taglines, positioning, objections and creative across nine languages. It is the only tool in this category a two-person comms team can buy on a card without a sales call. Read the evidence caveats below before you rely on it for anything consequential: this is a category where the marketing is well ahead of the proof, and Minds is among the more careful vendors in it rather than an exception to it.

For comms teams: pressure-testing your own thinking before you go to real people.

Best for

• Pressure-testing your own thinking before you go to real people
• Generating hypotheses and objections you had not considered
• Small teams with no research budget who need something better than guessing

Pricing

Free: 1 study and 3 audience answers/month. Individual $39 or €39/month for 500 synthetic responses. Team $79/seat/month with a two-seat minimum, 4,000 pooled responses per seat. Enterprise on application, adding calibration and validation studies and SSO.

Verified September 2026 from the vendor's own pricing page. The enterprise tier with calibration and validation is a materially different proposition to the $39 self-serve plan. Do not assume they are the same product.

Pricing last verified Sep 2026 – check vendor site for current rates.

What works well

+ Self-serve and affordable, which nothing else in this category is
+ Message and ad pretesting is a designed function, not a borrowed one
+ The vendor publishes a limits page and describes its own output as directional

Watch out for

- No peer-reviewed independent validation exists for this or any commercial synthetic-audience platform
- Independent research finds synthetic panels get the headline average roughly right while getting variance, subgroups and qualitative depth wrong
- Simulated respondents are systematically agreeable, so they are biased against telling you a message will fail
- Results shift with prompt wording and with model updates, so there is no version-locked instrument
- Vendor accuracy benchmarks replicate published surveys the model may already have been trained on

When to choose this

Choose Minds to generate hypotheses, surface objections and pressure-test a message cheaply before you take it to real people, never as the evidence that a message is safe to publish.

Who should look elsewhere

If the decision matters, use real people. Listen Labs runs AI-moderated interviews with a real human panel and is the more defensible purchase for anything consequential. If you need enterprise-scale simulation with calibration behind it, Aaru and Evidenza operate at that end, at prices to match. And if you simply want a second opinion on a draft, a general assistant costs less and makes fewer claims about what audiences think.

Key features

• Reusable AI personas built from your own documents, links and public data
• Interviews, group questioning and structured questionnaires
• Message, concept and advertising pretesting across nine languages
• Aggregate distributions plus individual simulated responses
• Published limits-of-simulation guidance, which is rare in this category

Governance & data

Data inputs Uploaded documents and links used to build personas, plus the questions and stimulus you test

Model training & controls Confirm whether uploaded brand and campaign material is retained or used for model improvement before uploading anything unreleased.

Admin, SSO & permissions Seat management on Team; SSO on Enterprise

Audit trail & approvals Studies and responses retained in the account

Risk notes The 2025 ICC/ESOMAR code distinguishes a person from a synthetic persona and requires synthetic output to be labelled as such. Reporting simulated results to a client, board or journalist as what audiences think would be a misrepresentation, and sits badly against CIPR and Global Alliance responsible-AI expectations. Say the simulation indicates, never audiences told us.

Governance information is based on publicly available vendor documentation. Verify with your vendor before procurement decisions.