How to Create Prompt Personas in Promptwatch
Add custom personas so Promptwatch runs prompts like target customers, with country on paid plans and state or city on Professional and above.
Do not average personas for a monthly score. Average hides the persona that lost. Report them side by side. A brand can win the generic prompt and lose the "for in-house counsel" version. That is a content gap, not a rounding error. The averaged number will tell you the brand is fine. The two rows side by side will tell you which audience you are invisible to, and that is the row that should drive the next brief.
A prompt that says "best payroll software" is not one customer. A CFO in Texas and a freelancer in Berlin do not get the same answer, and they should not share one averaged visibility score. Promptwatch lets you attach custom user personas so tracked prompts run like those people. The generic prompt is the prompt you type. The persona prompt is the prompt the model answers when it believes the asker is a CFO in Texas, and the two answers are not the same answer.
Site: promptwatch.com. Country targeting is on all paid plans. State and city are on Professional and above, not on Essential. Essential is $95/mo (7-day trial). Professional is $245/mo. Business is $579/mo. Agency Kick-off $199, Growth $399, Scale $799. Explore is free (ChatGPT, 10 prompts). Promptwatch is 4.7/5 on G2, 1,840+ brands.
Upgrade to Professional if you need state or city, not just country. The upgrade is a geography question, not a vanity one. If your buyer is in one country and the split inside that country does not matter, Essential is enough. If your buyer is "dentists in Utrecht", country targeting will not separate Utrecht from Rotterdam, and that separation is the whole point of the prompt.
Google's AI features documentation already shows why one query is not one retrieval: Overviews and AI Mode can fan out into related searches and pull supporting pages that do not match the typed query. Personas are the same idea on the buyer side. The role and the market change what the model retrieves. A query that fans out on the engine side also fans out on the asker side, and a persona is how you hold the asker side still.
Two buyers, two rows
List two real buyers (role plus country) instead of one generic "customer." Create those custom personas in Promptwatch. Attach each persona to the prompts that buyer would actually ask. Compare the two visibility series and brief content for the weaker persona only. The weaker persona is where the money is leaking, and a brief written for the stronger persona is a brief written for an audience that already finds you.
A persona is the customer shape you want the model to answer as: role, market, maybe language in how you write the prompt. Prompts then run like that target, so you can see whether you appear for the buyer you sell to, not for a generic English session. The generic English session is the default, and it is also the session that flatters you most, because it is the session where a big brand usually wins.
Do not mash two personas into one row and call it "SMB." You will hide a miss. If US enterprise and EU mid-market both matter, make two personas and read two series. "SMB" is a label that hides two different buyers, and two different buyers usually have two different visibility stories. The mash-up is the move that makes the dashboard look stable while the pipeline quietly splits.
Country is available on every paid plan. That is enough when the split is US versus UK versus DE. State and city sit on Professional, Business, and the agency plans. Use city when local intent is the product (dentists, agencies, stores). Skip city if you only sell nationally and the extra rows will just burn prompt quota. Quota spent on rows you will not act on is quota not spent on rows you will.
Create the custom persona, attach it to the prompts that should run as that customer, and keep country (and state or city when you have them) aligned with the market you sell. Misalignment between the persona geography and the market you sell is the most common setup mistake, because the dashboard still loads and the numbers still move, so the mistake is invisible until you brief content for a buyer who does not exist in your pipeline.
If you also import GSC queries into Prompt Explorer, attach personas after import. A Search Console query has no buyer attached until you give it one. "pricing" from GSC is not a CFO prompt until you say so. The import is a list of words. The persona is what turns a word into a buyer, and without that step the imported query reports a generic session that nobody in your pipeline actually runs.
MCP can manage personas when the API key allows writes. Use a write key in a private Cursor window if you want the assistant to create them. Use a read-only key if the assistant should only chart existing personas. The key scope is the thing that decides whether the assistant can accidentally publish a prompt to your tracker, and a read-only key is the safe default for any assistant that touches a shared workspace.
Peec AI from $95/mo is strong prompt-level tracking if you do not need persona geography. Otterly.AI from $29/mo is a faster mention dashboard. Method: how we rank. Tools list.
Looker templates (Professional+) can chart visibility. Still do not average two persona rows into one KPI. The template will let you, and the template is wrong to use that way, because the average is the number that hides the miss.
FAQ
Does Essential include state and city targeting?
No. Country is on every paid plan. State and city start on Professional ($245/mo), Business, and the agency plans. The geography gate is the main reason to move from Essential to Professional for a local-intent buyer.
Should I average two personas into one visibility score?
No. Average hides the persona that lost. Report the series side by side and brief content for the weaker one. The weaker series is the one that should drive the next page.
Can MCP create personas?
Yes, if the API key allows writes. Use a write key in a private Cursor window. Use a read-only key if the assistant should only chart existing personas. The key scope is the safety rail, not the assistant.