The most common questions. Honest answers.

In three steps: create an account with your email address, pick an object type (website incl. landing page, ad, email, logo, social post, product, Proposal Check, or Ask the Room), then select Personas and define your Review Criteria. Setup takes about ten minutes; generating the report takes a few more. After about 20 minutes, the finished PDF is ready.

No. Your email address is all it takes. A credit card is only requested when you upgrade to Pro or Business. No automatic plan conversion, no silent charges.

Yes, with the free plan: 2 Reviews per month, 2 Personas per Review — indefinitely. No trial period with an expiration date, no pressure to upgrade. You move up only when your volume grows.

German and English — end to end. The app's interface is fully available in both languages, and review output arrives entirely in your report language: persona voices, AI summary, quick wins, trade-offs, trust findings, and theme clusters. The persona library stays deliberately rooted in the German-speaking (DACH) market — that cultural depth is the point. Language itself is part of the model, too: personas with weak English say so openly when an English text only partly gets through to them — visible as a per-voice language marker in the report, and precisely the signal you want before deciding what to localize. The marker is the persona's own account of language friction, not a comprehension measurement.

About 20 minutes from upload to finished PDF report (median). Traditional UX research projects take 42 days on average (source: Dscout, survey of 300+ UX researchers).

Synthetic personas are AI-modeled stand-ins for real target groups: fictional profiles built on eight psychological layers (values, behavior, cognitive biases, and more). At Radical Personas, human experts co-design them and audit them for bias — they are not raw prompt output. They complement real user research; they don't replace it.

Essentially the same thing: both describe fictional, AI-generated persona profiles. “AI personas” emphasizes the technology; “synthetic personas” emphasizes the origin — artificially created rather than distilled from real interviews. Radical Personas uses both terms interchangeably; what matters is the depth of the modeling, not the label.

Good for fast pre-launch tests, hypothesis building, and decisions before budget is committed — feedback on websites, ads, emails, and product ideas in minutes instead of weeks. Not a substitute for real user research when you need final validation — and for niche or specialized audiences, the modeling becomes patchy.

No. The Nielsen Norman Group showed in 2024, for the category as a whole, that synthetic users miss the context-dependence of human behavior and turn patchy for niche or specialized audiences (nngroup.com/articles/synthetic-users). Our eight layers per persona are built against exactly that flattening — but building against it does not remove the limit. Synthetic personas complement real research; they don't replace it.

Stanford HAI (Park et al., 2024 — preprint, arXiv:2411.10109, n=1,052) found ≈85% replication accuracy, benchmarked against the test-retest reliability of human self-responses — that is, how consistently the same person answers the same question two weeks apart. Our internal comparison tests fall within that range; the methodology protocol is available on request.

Websites (including landing pages and user journeys across multiple pages — from Pro: up to 5 pages, from Business: up to 10, and as needed on Enterprise), ads, emails, logos, social media posts, products, Proposal Checks (reviews of proposals and quotes), and Ask the Room (one question put to a panel of Personas for a quick Sentiment Snapshot).

The Persona Finder reads what you want to have reviewed, derives the target customers from it and picks the best-fitting Personas for each target customer from the whole library — every Persona with a reason and the fit tag exact, close or substitute. If the library has nobody fitting for a target customer, it says so explicitly. Before it suggests anything it stops (“Does this look right?”): you deselect what does not fit or add a target customer of your own; up to that point nothing has been used. You also get a dissenting voice: a Persona likely to view your offering critically, with its objection in its own words — never pre-checked; if you take it along, it counts toward the Persona limit of your plan. The Finder is available on every plan, including Free. Daily quota per brand: Free 3, Pro 20, Business 60 analyses, Enterprise as needed; the first suggestion per brand is free.

Not in self-service right now. Over 200 library personas cover a broad range of Life-Worlds, cognitive biases, and decision styles — you select and filter. Custom-built Personas are available on request.

A Review is the evaluation of one object (for example a landing page, an email, or a logo) by one or more Personas. You get numeric Scores per criterion. Each Persona also delivers free-text answers in its own voice. An AI summary condenses the results across all Personas. Optionally, you can then chat directly with individual Personas.

Yes, from the Pro plan: user journeys across up to 5 pages (landing page plus follow-up pages). Business plan: up to 10. On Enterprise, journey depth follows your actual needs. Ideal for funnel analyses and checkout flows.

Our 8-layer engine combines the Big Five, Kahneman-style cognitive biases, distinct Life-Worlds, biography, and four further layers in a structured, internally validated interplay. ChatGPT gives you one one-dimensional answer per prompt. Details and a direct comparison at /en/vs-chatgpt.

A management summary (1 page), a detail report per Persona, and a summary across all Personas. Every PDF report includes Scores, free-text answers, and source references for the methodology.

Yes. From the Pro plan you can store your Workspace logo and company name — the generated PDF report comes out fully in your branding, without the Radical Personas logo. Ideal for agencies, consultants, and internal corporate reports that go straight to clients or stakeholders.

Synthetic market research tests concepts, copy, or designs on AI-generated stand-in personas before real respondents see them. 73% of market researchers already use synthetic responses in some form (Qualtrics 2025). It delivers fast directional signals — it does not replace a real sample.

No — they are not raw prompt output. Human experts co-design our library personas by hand, audit them for cognitive bias, and derive them partly from real survey data — across eight psychological layers. This human-plus-AI approach makes them more consistent and less generic than one-shot generated personas.

We run several language models in parallel and orchestrate their interplay with our own persona-layer engine. Details at /en/science.

Three layers are externally peer-reviewed and validated: psychology (OCEAN/Big Five — personality model validated across 56 nations, n=17,837 (Schmitt et al., 2007)), cognitive biases (Kahneman/Tversky, Nobel Prize 2002), and cultural context (Hofstede's cultural dimensions, established in cross-cultural research). Five further layers (biography, emotional state, behavior, anti-patterns, language) are our own synthesis — we check them internally for consistency against the three external layers. Details at /en/science.

Biography, psychology (OCEAN/Big Five), cognitive biases (Kahneman), emotional state, cultural context (Hofstede's cultural dimensions, value-based Life-Worlds), behavior, anti-patterns, and language. Details at /en/science.

Persona answers are consistent within a run — with an identical prompt and the same Persona, you get the same Scores. Between runs, free-text answers vary by about 8–12%. Methodology: internal comparison measurement (n=120 Review pairs, identical inputs 24 hours apart, metric: semantic similarity via embedding cosine distance). Numeric Scores typically vary less. The full measurement protocol is available on request.

Rosala and Moran (Nielsen Norman Group, 2024, nngroup.com/articles/synthetic-users) show, for the category as a whole: human behavior is complex and context-dependent, and synthetic users miss that complexity; for niche or specialized audiences the data turns patchy. Their conclusion: “Synthetic users cannot replace the depth and empathy gained from studying and speaking with real people.” We state this limitation openly and build around it: multi-persona Reviews, honest variance figures, no “AI replaces research” promise.

Yes — quarterly internal audits for typical large language model (LLM) biases (gender, racial, socioeconomic) plus a persona-stereotype check. An example from the Q1/2026 audit: personas from economically disadvantaged regions had value sets that were weighted systematically too conservatively — corrected through broader cultural-context sources and explicit correction guidelines. We also use multi-persona consensus (several Personas per Review), which makes stereotypical individual answers visible. Audit methodology and past reports are available on request, under NDA for Enterprise.

We don't run large-scale cross-validation studies ourselves — that would require the very panels we help complement. We anchor our approach in independent studies (Stanford HAI 2024: ≈85% replication accuracy relative to test-retest reliability; Qualtrics State of Synthetic Research 2025: 73% of market researchers already use synthetic responses; Nielsen Norman Group 2024). For Enterprise customers, we offer pilot validations against their own panels.

Hallucination is a known risk with large language models (LLMs). We mitigate it through: grounding each Persona in a fixed 8-layer profile (no free-form fact generation), multi-persona consensus per Review (a contradiction between Personas is visible, not hidden), and Score-range checks (implausible Score jumps get flagged). No system is error-free — which is why we work as a complement to real research, not a replacement.

In Austria and the EU. We comply with the General Data Protection Regulation (GDPR) and the Austrian Data Protection Act (DSG). Details are in our privacy policy.

No. Your content does not flow into the training of the underlying language models. Our contracts with the model providers prohibit any use of your data for model training.

While your account is active: indefinitely — responsibility for retention is yours, and you can export at any time during the term (data export as JSON in Settings, a PDF report per Review). Once the contract has ended for good, we delete the account data in line with the GDPR (Art. 17); invoice data is retained for 7 years (Austrian statutory retention requirement). Switching to the Free plan is not the end of the contract.

Yes. A data processing agreement (DPA under GDPR Art. 28) is a separate document that you conclude together with the terms during registration — on every plan, including Free. Full text in the application at app.radicalpersonas.com/avv; from Business we negotiate it individually. The subprocessor list (LLM providers, hosting) with jurisdictions is documented in the DPA. Change notifications go out 30 days in advance, with a right to object.

We pick the best available model per task so you get maximum review quality. Our contracts with every model provider rule out any use of your content for model training. Where contractually possible, we use the providers' EU endpoints; if processing takes place outside the EEA (for example via US endpoints), we secure it through EU Standard Contractual Clauses or the EU-US Data Privacy Framework. Which providers and processors we use — including their jurisdictions — is documented transparently in our privacy policy and the DPA. For Enterprise customers, an on-premises deployment is available on request — the data flow then stays entirely within your infrastructure.

Yes, from the Pro plan. The free plan is limited to 1 person (just you). Pro: up to 3 team members including the owner (owner + 2 invitees). Business: up to 6 (owner + 5 invitees). On Enterprise, we set the number together based on what you need. All members work in one shared account with a shared pool of Personas, Reviews, and Templates — practical for agency teams working on client projects together.

Yes, from the Pro plan. The free plan is limited to 1 Brand Profile. Pro: up to 5 parallel Brand Profiles; Business: up to 10. On Enterprise, we set up as many as you need. Each Brand Profile has its own Workspace — practical for agencies and consultants managing several client brands in parallel without switching accounts. The monthly Review quota is a shared pool across all Brand Profiles account-wide — not separate per brand.

Single sign-on (SSO — one central company login) is available from the Enterprise plan. Integrations with common corporate login systems such as Microsoft Entra ID, Okta, and Ping via SAML (the technical standard for single sign-on), plus automated user management (SCIM), are set up together with your IT team.

Yes. The Partner API lets you start Reviews from your own systems — for all eight object types of the app. The result is a Score and verdict per Persona plus a link into the app, where the full report lives; the API and the wizard run through the same code. The Persona Finder is available via the API as well, with the same daily quota as in the app. We issue keys from the Pro plan up, on request via /en/contact; there is no key on the Free plan. An MCP server for AI assistants — find Personas, start a Review, read the verdict, ask follow-up questions, share the report — is available on the Enterprise plan on request: add the server URL app.radicalpersonas.com/api/integrations/mcp with your key as a Bearer header to Claude Code, Claude Desktop or another MCP client; the assistant gets the core verdict, the full report lives in the app. Results are advisory pre-tests by synthetic, AI-generated personas and must be labeled as such when reused.

Email support is included in every plan. From the Pro plan, your requests are handled with priority. Enterprise customers additionally get a dedicated customer success manager (CSM) team. Target response times for critical incidents (P1 = priority 1) are tiered: Pro about 24 hours, Business about 8 hours, Enterprise about 4 hours; binding response times are agreed in the individual Enterprise SLA.

We run the service on modern, industry-standard infrastructure and aim for the highest availability we can deliver. We do not give a percentage uptime commitment in the standard plans. A binding uptime SLA with service credits is agreed individually for Enterprise. A public status page for incident history and real-time availability is in preparation.

Today: a PDF report on every plan (the basic version on Free, the full one with your own logo from Pro up), a share link for reports from Pro (the full report without login, with an expiry date and optionally a passcode) and the Partner API from Pro up with keys on request. An MCP server for AI assistants (Claude Code, Claude Desktop and other MCP clients) is available on the Enterprise plan on request. There are no native integrations for Figma, Notion, Slack or Zapier and no webhooks — to connect your own systems, use the Partner API. Requests welcome via /en/contact.

The free plan includes 2 Reviews per month with 2 Personas each — indefinitely, no credit card.

Yes. You can cancel at any time: a monthly subscription then ends with the current billing month, an annual subscription at the end of its term. No hidden fees, no minimum commitment beyond the booked interval.

Annual billing works out about 17% cheaper than paying monthly: you pay for ten months and use all twelve. The year is billed once in advance; your monthly quota stays the same.

The Enterprise plan has no tier caps. Volume, team size, Brand Profiles, and journey depth are set together based on what you actually need and written into your agreement. If usage becomes exceptionally heavy (for example automated bulk Reviews), our default is to reach out first before we restrict anything.

The full pricing FAQ lives at /en/pricing and covers payment methods (credit card, SEPA, invoice), VAT handling for Austria, Germany, and Switzerland, invoice downloads, the data processing agreement (DPA under GDPR), and how prices are index-linked to the consumer price index.

Anytime in your account area while the contract is active: the data export in Settings delivers your profile, workspaces, personas, Reviews, and chat histories as one JSON file, and every Review is also available as a PDF report — no export paywall. Save your data before the contract ends for good — after that, we delete the account data in line with the GDPR. Switching to the Free plan is not the end of the contract — your data stays.