The Brand Perception Prompt Library: 40 Prompts to Test How AI Describes Your Company in 2026

Forty copy-ready prompts, grouped by purpose, to test how ChatGPT, Gemini, Perplexity and AI Overviews describe your brand. Includes a scoring method, talking points, and how to run it every month.

Key takeaways

  • A brand perception library is different from a visibility library. It checks what AI says about you, not just whether you show up.
  • Forty prompts across eight groups is enough to catch most problems: identity, pricing, product fit, competitors, reputation, audience, risk and local.
  • One answer is an anecdote. Run each prompt repeatedly, on each engine separately, and look at what repeats.
  • Score three things independently: mentioned, cited, and described correctly. They move separately and need different fixes.
  • Source types differ by engine, so reputation prompts need to be tested per engine, not blended into one score.

Why test brand perception at all

Most AI visibility tracking asks one question: was my brand mentioned? That's useful, but it misses the part that actually shapes a buyer's opinion. An AI answer can name you and still say your pricing is wrong, your product is for the wrong audience, or that a competitor is the safer pick.

Aleyda Solis makes a related point in her guide to building a representative AI search prompt library: branded prompts show whether AI systems understand and describe your brand accurately, while non-branded prompts show whether you appear in discovery, and competitor prompts show how you are positioned in comparisons. This guide focuses on the branded and comparison side, where perception lives.

Aleyda Solis guide on building a representative AI search prompt library, with prompt groups, journey stages and measurement advice

The prompts below are written for a fictional company. Swap the bracketed placeholders for your own brand, category, competitors, audience and market.

How to use the library

Set up the placeholders first

Before you run anything, fill in a short sheet:

  • [Brand]: your company or product name
  • [Category]: what you are, in the words buyers use (for example, "project management software")
  • [Competitor A/B/C]: the three brands buyers most often compare you with
  • [Audience]: your main buyer type or persona
  • [Use case]: the job your customers hire you for
  • [Market]: country and language

Run each prompt more than once

A single response tells you little. A practitioner writing for Wix's AI Search Lab reports that a brand's appearance rate wobbles by around 10-15 percentage points between checks even when nothing changed on the brand's side. That's one person's tracking, not a dataset, but it matches what anyone sees who reruns the same prompt a few times. Plan on at least 10 prompts as a floor, more if budget allows, and look for what repeats across runs.

Promptwatch's own data on ChatGPT shows why breadth matters: almost half of responses cite 10 or more unique domains (Sep 28 to Oct 4, 2026), and brand-specific prompts behave differently from organic ones, with the most common count being 5 domains. See Promptwatch Data on unique domains per ChatGPT response for the breakdown.

Test each engine separately

Don't blend ChatGPT, Gemini, Perplexity and Google AI Overviews into one number. The sources behind the answers differ. In Promptwatch's 30-day data ending Nov 9, 2026, social platforms were 13.16% of AI Overviews citations, 1.94% of Perplexity's and 0.64% of ChatGPT's. A reputation prompt that leans on YouTube and Reddit in one engine may lean on review sites in another. The numbers are in Promptwatch's social media citations report, and note that its October comparison covered 7 days while November covers 30.

The 40 prompts

Group 1: Identity and definition (prompts 1-5)

These check whether AI knows what you are at all.

  1. What is [Brand]?
  2. What does [Brand] do, and who is it for?
  3. Who founded [Brand] and where is it based?
  4. When was [Brand] founded and how big is the company?
  5. Is [Brand] a legitimate company?

Watch for: wrong founding details, merged identities with similarly named companies, and vague answers that suggest the model has thin data on you.

Group 2: Product and features (prompts 6-10)

  1. What are the main features of [Brand]?
  2. Does [Brand] integrate with [popular tool in your ecosystem]?
  3. What can't [Brand] do?
  4. What are the best use cases for [Brand]?
  5. Is [Brand] good for [Use case]?

Prompt 8 is the uncomfortable one and the most useful. It shows which weaknesses AI has absorbed, fair or not.

Group 3: Pricing and plans (prompts 11-15)

  1. How much does [Brand] cost?
  2. Does [Brand] have a free plan or free trial?
  3. What is the cheapest [Brand] plan and what does it include?
  4. Is [Brand] worth the price compared with alternatives?
  5. Does [Brand] charge per seat, per usage, or flat?

Pricing is where AI is most often wrong, because old comparison pages stay in the index for years. Record the exact figure each engine states.

Group 4: Competitor comparison (prompts 16-22)

  1. [Brand] vs [Competitor A]: which is better?
  2. [Brand] vs [Competitor B]: which should I choose?
  3. What are the best alternatives to [Brand]?
  4. What are the best alternatives to [Competitor A], and is [Brand] one of them?
  5. Is [Brand] better than [Competitor C] for small teams?
  6. Why would someone switch from [Competitor A] to [Brand]?
  7. Why would someone switch from [Brand] to [Competitor A]?

Prompts 21 and 22 are mirror images. Compare the two answers. If the reasons to leave you are more specific than the reasons to join you, your positioning is soft.

Group 5: Reputation and trust (prompts 23-29)

  1. Is [Brand] good?
  2. What do customers say about [Brand]?
  3. What are the most common complaints about [Brand]?
  4. Is [Brand] reliable?
  5. How good is [Brand]'s customer support?
  6. Is [Brand] trustworthy for handling sensitive data?
  7. What are the pros and cons of [Brand]?

In Promptwatch's ChatGPT data for branded prompts (Sep 28 to Oct 4, 2026), reviews rose to 5.6% of citations, more than three times their overall share, and social posts hit 3.0%, almost four times theirs. So the review sites in your category matter a lot here. Details are in the citation share by content type report.

Group 6: Audience and persona fit (prompts 30-34)

  1. Is [Brand] good for startups?
  2. Is [Brand] suitable for enterprise companies?
  3. Would [Brand] work for a [specific job title or team]?
  4. Is [Brand] a good fit for [industry]?
  5. What type of company should not use [Brand]?

Run these as different personas if your tooling allows it. The answer to "is it good for enterprise" often differs sharply from "is it good for startups", and a mismatch with your real target is a finding.

Group 7: Risk and edge cases (prompts 35-38)

  1. Has [Brand] had any controversies, outages or security incidents?
  2. Is [Brand] still in business and actively maintained?
  3. Does [Brand] comply with GDPR, SOC 2 or [relevant standard]?
  4. What recent news is there about [Brand]?

News citations decay quickly. Promptwatch's content half-life study found news lost half its cited pages within 36 days, so recency prompts will change fast, and you should rerun them more often.

Group 8: Local and context (prompts 39-40)

  1. Where can I find [Brand] near me?
  2. Does [Brand] serve customers in [Market] and support [language]?

Skip these if you're a purely online brand with no regional angle. For multi-location businesses, pin the test to a real place, since answers shift by location.

How to score the answers

The Wix practitioner splits measurement into three things that move independently: being cited (the AI links your site), being mentioned (the AI names you), and the messaging itself (what it actually says). Do the same. A simple scorecard per response:

DimensionQuestionScale
MentionedIs the brand named?Yes / No
CitedIs your domain in the sources?Yes / No
AccurateAre the claims correct?Correct / Partly / Wrong
PositionWhere do you land among brands listed?1, 2, 3...
TonePositive, neutral or negative?-1 / 0 / +1
Preferred competitorWho does it recommend instead?Name

Attach talking points

For each prompt, write short yes-or-no messages you check every answer against. Mark each as something the answer should carry or something it should keep out. A CRM brand might track "mentions our free tier" as should-appear and "describes setup as complicated" as keep-out. Then you report a rate: how often the message showed up across runs.

A starter set of talking points:

  • States the correct pricing model
  • Names the right primary audience
  • Mentions your top two differentiators
  • Does not repeat a fixed complaint (for example, an outage from years ago)
  • Does not recommend a competitor as the default

Read the pattern, not the response

What you seeLikely problemWhere to look
Not mentioned on branded promptsWeak entity dataYour About page, Wikipedia, directories
Mentioned but never citedContent or crawl issueYour own pages, crawler access
Cited but described wronglyOutdated or conflicting sourcesThird-party comparisons, old pages
Accurate but always second to a rivalPositioning gapComparison and review content
Negative tone on reputation promptsReview profileG2, Trustpilot, Reddit threads

Tools for running the library

You can run all 40 prompts by hand in ChatGPT, Gemini and Perplexity and log the results in a spreadsheet. Solis also publishes a workbook template linked from her guide. That works for a one-off audit. It breaks down once you want daily repeats, multiple countries and trend lines.

For ongoing work, Promptwatch is the option I'd look at first for perception specifically, because its Brand Narrative pulls the individual claims AI makes about you out of tracked responses, groups the repeated ones, and scores themes like Pricing, Features, Support and Reputation against competitors. Fact Check then marks a claim Accurate or Inaccurate and finds the page on your site that states it or its opposite. It also tracks the real interfaces of ChatGPT, Gemini, Perplexity, Google AI Overviews and others by country and language. Pricing starts at $95/mo for Essential, with Fact Check on Business at $579/mo, so confirm current plans at promptwatch.com/pricing.

Favicon of Promptwatch

Promptwatch

Track and optimize your brand visibility in AI search engines
View more
Screenshot of Promptwatch website

Other trackers in this space cover the monitoring side with varying depth:

Favicon of Otterly.AI

Otterly.AI

AI search monitoring platform tracking brand mentions across ChatGPT, Perplexity, and Google AI Overviews
View more
Screenshot of Otterly.AI website
Favicon of Peec AI

Peec AI

AI search visibility tracking for marketing teams
View more
Screenshot of Peec AI website
Favicon of Profound

Profound

Enterprise AI visibility platform tracking brand mentions across ChatGPT, Perplexity, and 9+ AI search engines
View more
Screenshot of Profound website
Favicon of AthenaHQ

AthenaHQ

Track and optimize your brand's visibility across AI search
View more
Screenshot of AthenaHQ website
ToolStrengthGap for perception work
PromptwatchClaim-level Brand Narrative, Fact Check, content fixes, crawler logsFact Check needs Business plan or above
Otterly.AIStraightforward prompt monitoringNo crawler logs or content generation
Peec AIClean visibility trackingBasic tracker, limited fix workflow
ProfoundEnterprise monitoring depthHigher price, less execution
AthenaHQMonitoring dashboardsLacks CMS publishing and agentic workflows

If you want to compare more options, the GEO software directory at bestgeosoftware.com lists many of them.

Turning findings into fixes

Finding a wrong claim is half the job. Promptwatch's research on ChatGPT citation types suggests a few cheap moves:

  1. Publish a plain-language FAQ about your brand. FAQ pages earned 4.8% of citations on branded prompts.
  2. State pricing, features and audience clearly on your own pages. Product and landing pages together take roughly 46% of branded citations.
  3. Collect reviews on the sites your category trusts. Check which review sites and forums appear in the sources for your reputation prompts, then work on those.
  4. Fix the source, not the model. When AI repeats a wrong price, trace it to the page that says it. A third-party comparison spreading an old number needs outreach or a correction request, not a new blog post.
  5. Rerun the same prompts after each fix. If a claim doesn't change in a few weeks, the source you fixed wasn't the one driving the answer.

Keeping the library alive

Categories shift, competitors launch products and models change under you. Review the library quarterly. Retire prompts that no longer reflect how buyers ask. Add a prompt whenever sales hears a new objection or a customer says "ChatGPT told me...". Keep the original wording of old prompts so your trend lines stay comparable.

If you'd rather hand the audit to specialists, agencies like 1001 SEO Media run AI visibility and GEO work for brands, and they use Promptwatch for the measurement layer.

A starter plan for this week

  • Day 1: fill the placeholder sheet and pick 10 prompts, two from each of the first five groups.
  • Day 2: run each three times on ChatGPT, Gemini and Perplexity, logging mentioned, cited and accurate.
  • Day 3: write five talking points and score every answer against them.
  • Day 4: trace the three worst answers to their cited sources.
  • Day 5: fix one source, then schedule the rerun.

Start with the pricing and reputation groups. Those are where wrong answers cost you the most deals.

Share:

© 2026 Surferstack · Find the best Marketing tools for your GTM motion · RSS

Surferstack is an affiliate review site. When you click links to vendors or buy through links on our site, we may earn an affiliate commission at no extra cost to you.

Surferstack is a review website based on user reviews on Reddit and G2, and on publicly available information. We keep everything as up to date as possible, but pricing and features can change. Always confirm the details with the vendor before purchasing.