How to Create Prompt Personas in Promptwatch
We rank Promptwatch first for prompt personas that run like target customers without averaging those rows together.
You cannot get city-level persona checks on Essential. Essential is country-level. State and city start on Professional and on agency plans. Do not roll every persona into one visibility score either. Keep persona rows separate. An average hides the customer you are losing.
The reason we lead with these two limits is that they are the two mistakes buyers make with personas. The first is assuming personas come with city targeting on the cheap plan. They do not. Essential is country-level, so a persona on Essential is a persona asked from a country, not from a city. The second is averaging persona rows into one score, which produces a number that hides the persona that is losing. Both mistakes come from treating a persona as a label instead of a separate series.
A generic "best CRM" prompt is not what a finance lead in Austin types. We rank Promptwatch first for prompt personas because you store the customer type on the prompt and read each persona's answers as their own rows. Country-only mention tools sit in the directory. Review: Promptwatch. Product: promptwatch.com.
The finance lead example is the one to sit with. A finance lead in Austin does not type "best CRM." They type something closer to "best CRM for a 50-person finance team with SOC 2 requirements." That prompt is a different question from the generic one, and it gets a different answer. A persona stored on the row is what turns the generic prompt into the finance lead's prompt. The persona is the customer type, and the prompt is the question that customer asks.
Personas do not change how Google builds an Overview. Google's AI optimization guide is still the official read on helpful content for those answers. The persona row changes who the stored prompt is for. It does not add a fake engine, and it does not rewrite Google's rules.
The reason to say this out loud is that buyers sometimes assume a persona changes what Google does. It does not. A persona changes who the stored prompt is for, which changes what the model answers, because the model answers the prompt it reads. It does not change Google's rules for building an Overview, and it does not add an engine the product does not have. The persona is a control on the row, not a switch on the platform.
Country tracking is on paid plans. State and city tracking starts on Professional ($245/mo) and on agency plans (Kick-off $199, Growth $399, Scale $799). Essential is $95/mo (country-level, MCP and API). Explore is free (10 ChatGPT prompts). Business is $579/mo. 4.7/5 on G2, 1,840+ brands.
Brief it like a researcher
Name two real buyers (role and constraint) and write one prompt each would type. Create those personas in Promptwatch and attach them to the prompts. Write the persona the way you would brief a researcher: role, industry, constraints, and the wording that person uses. The stored run is then closer to that customer's answer than a single intern string reused for every market.
The brief-it-like-a-researcher framing is the one to use. A researcher brief has a role, an industry, a constraint, and a wording. "Finance lead at a 50-person firm, SOC 2 required, asks about CRM for regulated finance" is a brief. "Intern" is not a brief. The persona field is where the brief goes, and the prompt is the question that brief would type. The stored run is then the answer that customer would get, not the answer a generic string gets.
We will not invent a persona wizard with extra screens. The sourced rule is simple: prompt like target customers, keep persona rows separate.
If you also need geography, pair the persona with location. Country is enough for "US vs UK." A clinic chain that cares about Dallas vs Houston needs Professional or an agency plan. A SaaS brand comparing the US and Germany can stay on Essential for country. Run them on Essential if country is enough, or Professional if you need a city. The Austin finance lead is a persona plus a city. The UK intern is a different persona plus a country. Those are four facts on two rows, not one blended score.
The geography pairing is what turns a persona into a located persona. A persona without geography is a customer type asked from the default market. A persona with country is that customer type asked from a country. A persona with city is that customer type asked from a city. The Austin finance lead needs the city, because Austin and Dallas answer differently. The UK intern needs the country, because the UK is the market. Those are two different rows with two different location settings, not one row.
Paid engines remain daily real UI: ChatGPT, Gemini, Claude, Perplexity, Grok, Llama, DeepSeek, Mistral, Copilot, AI Overviews, and AI Mode. Personas change who the prompt is for. They do not add a fake engine.
Read the two series. Do not mash them.
Do not average "CFO persona" and "intern persona" into one visibility number. Those are different questions. Report them as separate series, the same way you would not average branded and unbranded Google queries without saying so. Read the two answer sets side by side. Add a third persona only after the first two changed a content decision.
The reason not to average is that an average hides the loser. If the CFO persona sees you and the intern persona sees a competitor, a blended score reads as a tie. The meeting needs to know the intern persona is losing, because that is the ticket. Averaging branded and unbranded Google queries is the analogy: you would not average those without saying so, because they answer different questions. Persona rows are the same. They are different questions and they belong in different columns.
MCP can read personas and, if the key allows writes, create them. Use a read-only key when the chat should not edit the roster.
| Product | Personas we can rank | Notes |
|---|---|---|
| Promptwatch | Personas on prompts; state/city on Professional, Business, or agency | Do not average persona rows |
| Otterly.AI | 65+ countries, mention UI | $29; 4 engines; Gemini add-on |
| Peec AI | Prompt scores | $95, 3 models |
| Profound Starter | ChatGPT prompt set | $99/mo annual, ChatGPT-only |
Read the table as one row that stores personas on prompts and three rows that do not. Promptwatch is the row where the persona lives on the prompt and the location can go down to city. Otterly's 65+ countries is coverage, not a persona ledger, which means it tells you where the prompt ran but not who the prompt was for. Peec scores prompts but does not store the customer type. Profound Starter is a ChatGPT prompt set, which is one engine and no persona field. The persona ledger is the feature that answers "run the prompt as this customer," and only one row has it.
Otterly's country list is coverage, not this persona ledger. We rank Promptwatch first when the job is "run the prompt as this customer."
FAQ
Can I run city-level persona checks on Essential?
No. Essential is country-level. State and city tracking starts on Professional ($245/mo) and on agency plans.
Should I average two personas into one visibility score?
No. Keep persona rows separate. An average hides the customer you are losing. Report them as separate series.
Do personas add extra engines?
No. Paid engines remain the daily real-UI set. Personas change who the prompt is for. They do not add a fake engine.