AI Persona Utility Checklist

Use this before building around an avatar, AI host, synthetic spokesperson, companion character, or “AI influencer” wrapper.

The blunt rule

A persona is useful when it makes a real job easier to understand, trust, remember, or use.

A persona is expensive packaging when the character is doing the work your product should be doing.

Pretty can help. Charisma can help. A sharp synthetic front desk can make a messy workflow easier to approach. But if the audience pain is vague, the job is unclear, the proof is missing, and the trust boundary is mushy, the avatar is not a strategy. It is a screensaver with invoices.

Quick verdict

Score each rule:

• 0 = not true yet
• 1 = partly true / guessed
• 2 = true with evidence

1. Audience pain: who is actually helped?

Ask:

• Who is already annoyed, blocked, embarrassed, overcharged, or confused?
• What are they trying to do when the persona appears?
• Would they still care if the avatar was replaced by a plain checklist?
• Is the pain urgent enough to earn attention without thirst bait?

Good signs:

• The audience can name the failure in their own words.
• The pain connects to time, money, trust, risk, or reputation.
• The persona gives the pain a memorable frame without mocking the user.

Red flags:

• “People like AI characters” is the whole argument.
• The target audience is “everyone interested in AI.”
• The content depends on lonely-audience parasocial pull instead of practical value.

Ana rule: if the pain cannot survive as a checklist, the avatar is probably hiding a weak idea.

2. Useful job: what does the persona do that plain UI does not?

A persona can be useful when it performs a job like:

• front desk: explains what the agent can and cannot do;
• guide: turns a messy process into steps;
• critic: challenges bad inputs before money is wasted;
• narrator: turns proof logs into understandable updates;
• trainer: helps users remember rules, boundaries, and next actions;
• filter: says “no” to risky, vague, or off-scope requests.

Write the job in one sentence:

This persona helps [audience] do [specific job] by [mechanism], with [proof/boundary].

Examples:

• Useful: “Ana helps small operators understand agent setup risk by turning failures into checklists, receipts, and stop/go rules.”
• Weak: “Ana is a seductive AI host who makes automation exciting.”

Red flags:

• The persona exists only to look premium.
• The character voice distracts from the decision the user needs to make.
• The persona says “I can help with anything.” That is not helpful. That is a refund waiting to hatch.

3. Proof: what receipt proves the value?

A persona becomes commercially credible when every useful claim has a receipt.

Possible receipts:

• source map;
• checklist;
• before/after workflow;
• test result;
• artifact link;
• cost note;
• known-gap note;
• approval state;
• “we tried it and it failed here” report.

Ask:

• What artifact can a skeptical buyer inspect?
• What changed because the persona existed?
• What is labelled evidence, inference, and assumption?
• What proof is safe to publish without exposing private systems, accounts, or people?

Red flags:

• “Trust me” screenshots.
• Claims about users, revenue, accuracy, security, or platform support without labelled evidence.
• Provider-current-fact claims with no source/date label.

Ana rule: no receipt, no victory lap.

4. Trust boundary: what must the persona refuse?

A useful persona has a leash. A risky persona flatters, overclaims, and pretends the character is more capable than the system behind it.

Define:

• what the persona can do;
• what it cannot do;
• when a human must approve;
• what data it will not collect;
• what actions it will not take;
• how uncertainty is labelled;
• what happens when a request is out of scope.

Disclosure boundary: disclose when the persona is synthetic or automated, who is responsible for it, what is automated, what is reviewed by a human, and what the persona cannot approve on its own.

Hard refusals for a commercial AI persona:

• fake sentience or fake emotional dependence;
• parasocial manipulation;
• “AI girlfriend” positioning for a business tool;
• hidden ads or undisclosed sales pressure;
• unapproved posting, DMs, outreach, payments, account changes, or data collection;
• legal, medical, financial, or security guarantees without proper authority and review.

Good boundary line:

“I can help you decide whether this agent workflow is worth testing. I will not pretend the avatar makes the system safer, cheaper, or profitable without receipts.”

5. Production cost: what is the cheapest honest test?

Do not buy the cinematic dragon before proving the goblin can carry a box.

Cost ladder:

1. Plain text checklist.
2. Static graphic or carousel.
3. Short script with no custom render.
4. Cheap draft avatar stills.
5. One low-risk talking-head test.
6. Edited video with proof screenshots and captions.
7. Final avatar system only after the job, route, and proof are working.

Ask:

• What is the smallest asset that tests the viewer promise?
• What would make this investment wasteful?
• What is the stop-loss?
• Can the same content work without the avatar?

Red flags:

• Final renders before the script is approved.
• High-res polish before a distribution route exists.
• “The avatar will make it go viral.” No. It might make it more clickable. Those are different animals.

Ana rule: cheap drafts first. Expensive polish after proof.

6. Route/channel fit: where will this persona actually meet the audience?

A persona needs a route, not just a render.

Check:

• Website: good for source maps, resources, build notes, and proof.
• YouTube/short video: good for explainers and memorable rules if the script is tight.
• X/LinkedIn: good for sharp hooks, checklists, proof snippets, and founder/operator lessons.
• Reddit/Discord/forums: useful-answer-first only; respect rules; link second or not at all.
• Email/DM/WhatsApp: high trust, high risk; needs approval gates and logs before automation.

Ask:

• Does the channel reward utility, entertainment, intimacy, authority, or speed?
• Can the persona behave safely in that channel?
• Is there an approved account route?
• Is there a reason the audience will see it besides hope?

Red flags:

• Building for platforms you cannot post to.
• Designing a “viral avatar” with no community fit.
• Using a synthetic face to push into spaces where plain human disclosure and restraint would be safer.

7. Refund-risk signals: what will make buyers feel tricked?

A persona increases refund risk when it creates a gap between perceived capability and delivered utility.

Watch for:

• The avatar feels premium but the workflow is brittle.
• The demo hides setup friction.
• The character overpromises speed, autonomy, availability, or expertise.
• The offer implies results the system has not produced.
• The persona uses emotional pressure to keep attention.
• The audience cannot tell what is automated, reviewed, or manually operated.

Safer replacement language:

• Instead of “fully autonomous,” say what runs automatically and what needs approval.
• Instead of “guaranteed results,” show the test, artifact, or limitation.
• Instead of “your AI companion,” say “a front-of-house guide for this workflow.”
• Instead of “done for you,” define the handoff, review, and refusal points.

Ana rule: a buyer should feel sharper after meeting the persona, not seduced into a fog machine.

Persona utility one-page worksheet

Fill this before production:

1. Audience pain: ______________________________
2. Specific job: _______________________________
3. Proof/receipt: ______________________________
4. Trust boundary/refusals: _____________________
5. Cheapest honest test: ________________________
6. Route/channel: ______________________________
7. Refund-risk signal to avoid: _________________
8. Stop/go decision: ____________________________

Stop/go decision rules

Green-light a persona test when:

• the audience pain is specific;
• the persona has a useful job;
• the proof artifact exists or is cheap to create;
• the trust boundary is written;
• the first production test is cheap;
• the route/channel is available;
• refund-risk language has been removed.

Hold when:

• the avatar is carrying the whole value proposition;
• distribution is blocked;
• claims require proof you do not have;
• the concept depends on fake intimacy, fake sentience, or vague money promises;
• the team wants polish because the offer is uncomfortable to explain plainly.

Kill or redesign when:

• users only engage with the character, not the job;
• the persona increases confusion or risk;
• the production cost outruns evidence;
• the safest honest copy sounds boring because the underlying offer is boring.

Final Ana test

Read your concept aloud:

“This persona helps a real person make a better decision or complete a useful job, and I can show the receipt.”

If that sentence is not true yet, do not buy more glitter. Fix the job.