25 Competitor Analysis Prompts for ChatGPT, Claude and Gemini (2026)

Most competitor analysis prompts floating around stop at one line: "Analyze my competitor, [brand name]." You can guess the output — a rewritten encyclopedia entry, three strengths, three weaknesses, and nothing you could take into a Monday planning meeting. The model isn't the problem. The prompt is, because it never says who you are, what decision you're making, or what shape the answer needs to be in.

What follows is a library of 25 prompts, each matched to one of three models — ChatGPT, Claude, or Gemini — since they're strong at different jobs. Every prompt is written out in full, ready to copy, with placeholders in [square brackets] to swap.

These are the prompts we use at Intura when we map a competitive landscape before building an AI search strategy for a brand; the ones that made this list are the ones that survived after we cut every version that produced generic output.

Why do generic competitor analysis prompts always return shallow answers?

Without context, a model can only return the average of everything ever written about your category — and averages never help you decide anything. It doesn't know whether you're rewriting a pricing page or weighing entry into a new segment, so it picks the answer that's safe for both.

Three failures repeat every time the prompt is too short:

• Ownerless findings — the same strengths and weaknesses would be true for any brand in the category.
• Untraceable numbers — pricing, user counts, and launch dates arrive with total confidence and zero sources.
• Unusable output — seven paragraphs of prose when what you needed was a table you could paste straight into the strategy doc.

What makes a competitor analysis prompt actually work?

Working prompts carry five things: a role, the decision at stake, the raw material, the output format, and an honesty rule. Drop any one and quality falls immediately — and the two people drop most often are the last two.

ChatGPT, Claude, or Gemini — which one should you use for competitor analysis?

Choose by task type, not by favourite: use a model with live web search to gather facts, and a long-context model to dissect material you've already collected.

Task type	What we use	Why
Gathering current facts (pricing, launches, news)	Gemini or ChatGPT with search on	Answers stay tied to pages you can open and check yourself
Dissecting long material (articles, transcripts, hundreds of reviews)	Claude	Long context, and its summaries tend to keep the specific details
Writing memos, battle cards, recommendations	Claude or ChatGPT	Both hold to strict formatting instructions
Checking which brands AI names in your category	All three, run separately	Each names a different set — one model is not a sample

One working rule saves more time than any model choice: never ask a model to recall data you could paste. Pasting 40 real reviews produces sharper analysis than asking it to "analyze reviews of product X".

Prompts 1-5: How do you map who you actually compete with?

Start with the map, because a wrong competitor list sends every later analysis in the wrong direction. These five usually produce a wider set of rivals than the one in your internal deck.

1. Three-layer competitor map

Best in: Gemini or ChatGPT with web search on.

You are a market research analyst for [brand name], which sells [product/service] to [target audience] in [market]. Build our competitor map in three layers: (1) direct competitors selling the same solution to the same audience, (2) indirect competitors solving the same problem a different way, (3) non-consumption alternatives, meaning what the audience does if they buy nothing at all. For every name, give the reason it belongs in that layer, your source, and your confidence level (high/medium/low). Clearly flag anything you cannot verify.

2. One-page competitor profile

Best in: ChatGPT or Gemini with search on.

Build a one-page profile of the competitor [competitor name] ([URL]). Structure it as: value proposition in one sentence, primary customer segment, product lines and published pricing, visible acquisition channels (SEO, paid, social, marketplace, partnerships), the three differentiation claims they make on the homepage, and three things they clearly do not offer. Close with a section titled "What I could not confirm" listing the questions that need manual research.

3. Competitors that aren't on your radar

Best in: ChatGPT — the consumer role makes the output more realistic.

Play the role of [customer profile, e.g. a 32-year-old mother in Bandung looking for oily-skin skincare under Rp150,000]. Write the 15 searches or questions you would genuinely type into Google, TikTok, or ChatGPT before deciding to buy. For each one, name the brands or sources most likely to come back as the answer. Then list every brand that appeared more than once but is missing from our competitor list: [your competitor list].

4. Twelve-month competitor timeline

Best in: Gemini with search on.

Build a timeline of [competitor name]'s moves over the last 12 months using public sources only: product launches, pricing changes, major campaigns, homepage message changes, channel expansion, and press coverage. Present it as a table with columns for month, event, source, and implication for [our brand name]. If you cannot find a source for an event, leave the event out entirely.

5. Check which brands AI names in your category

Best in: run it separately in ChatGPT, Claude, and Gemini, then compare.

Answer the following as you normally would, without trying to work out which brand I represent. If someone in [market] asks "[category question, e.g. best point-of-sale app for small businesses]", which brands would you name, in what order, and based on what kind of information? Repeat for these five question variants: [list 5 questions]. Afterwards, explain the pattern: what type of source shaped your answers most.

Prompts 6-10: How do you dissect a competitor's content and visibility?

This block answers the question content teams ask most: why does their page get pulled and ours doesn't. All five work best when you paste the page content rather than the URL.

6. Structural teardown of a competitor article

Best in: Claude.

Here is the full text of a competitor article: [paste article]. Take apart its structure and answer five things: (1) what question is answered in the first 100 words, (2) list every heading and the question each one answers, (3) which passages could be quoted whole by an AI engine because they stand alone without prior context, (4) which claims carry a source and which do not, (5) three concrete weaknesses we could beat. Do not review the writing style; focus on structure and evidence.

7. Content gap analysis

Best in: Claude — long title lists need the context room.

Here are our article titles: [paste list]. Here are the article titles from three competitors: [paste list]. Find three groups: (1) topics they cover and we don't, (2) topics we cover but where their version goes deeper, (3) topics nobody covers that [audience description] clearly asks about. Rank group three by potential impact and justify each ranking individually.

8. Buyer question map by stage

Best in: ChatGPT.

For the [category] category in [market], write the 30 questions a prospective buyer is most likely to type, grouped into awareness, consideration, and decision stages. For each question, guess which of [competitor list] has most likely already answered it well, and mark the questions nobody has answered. Label your guesses as guesses — do not present them as data.

9. GEO audit of a competitor page

Best in: Claude.

This page keeps showing up as a source in AI answers about [topic]: [paste competitor page content]. Explain technically why this page is easy for AI engines to cite, examining: heading structure, how directly the first sentence of each section answers its heading, the presence of definitions, tables, numbered lists, sourced data, and specific named entities. Then compare it against our page: [paste our page content]. Close with a fix list ordered by expected impact.

10. Content-to-product funnel map

Best in: Claude or ChatGPT.

Here is a competitor's path from blog article to product page: [paste URLs and a summary of each page]. Map their funnel: what promise is made at each step, what friction they removed, where they ask for visitor data, and what offer they use to push conversion. Point to the single weakest link in that funnel and explain how [our brand name] could exploit it.

Prompts 11-15: How do you read a competitor's positioning, messaging, and pricing?

Competitor positioning can almost always be reconstructed from public copy — the homepage, the pricing page, and the about page are enough. These five turn that copy into something you can lay side by side with your own.

11. Reconstruct their positioning statement

Best in: Claude.

Here is a competitor's homepage, pricing page, and about page copy: [paste]. Reconstruct their positioning statement in this format: for [who], who [problem], [product] is a [category] that [key benefit], unlike [alternative] because [differentiator]. Then show which parts of that positioning are genuinely supported by the product they describe, and which parts are marketing claim only.

12. Test their message against customer language

Best in: Claude.

Compare three inputs: (1) the competitor's core messaging [paste copy], (2) their customers' complaints and praise [paste reviews], (3) the words the audience actually uses when discussing this problem in forums and comment sections [paste]. Show where their messaging misses the customer's language, then write three message angles for [our brand name] that fill the gap. Use the customer's vocabulary, not corporate vocabulary.

13. Pricing structure comparison

Best in: ChatGPT.

Here is the pricing from [3-5 competitors]: [paste plan names, prices, limits, and features]. Build a comparison table with columns for plan name, price, usage limits, differentiating feature, and the buyer type it serves best. Then answer three questions: which price band is most crowded, which band is empty, and what the concrete risk is if [our brand name] moves into the empty one.

14. Map the objections they never answer

Best in: Claude.

Here is a competitor's FAQ page and pricing page: [paste]. List the buyer objections they clearly anticipate and how they answer each. Then list the objections that obviously exist in this category but that they never touch. For each untouched objection, write one honest paragraph answering it from [our brand name]'s point of view, without disparaging the competitor.

15. Brand voice analysis and the open space

Best in: Claude.

Analyze the competitor's writing style from this material: [paste 5-10 samples of copy or captions]. Describe the formality level, average sentence length, use of technical terms, how they address the reader, and three sentence patterns that repeat. Then propose a voice position for [our brand name] that sounds distinct in this category without feeling forced, and write one sample paragraph in that voice.

Prompts 16-20: What can you read from a competitor's social and customer voice?

A competitor's comment section is the most underused free market research there is. These five run on data you collect yourself — captions, metrics, reviews — not on the model's memory.

16. Which social formats actually work for them

Best in: Claude.

Here are a competitor's 20 most recent posts with engagement counts: [paste captions and metrics]. Group them by content format (education, product, entertainment, user-generated, promo). For each group, calculate average engagement and describe the opening-line pattern they use. Close with three conclusions: which format genuinely works for them, which they keep doing out of habit, and one format they have never tried.

17. Cluster the customer voice

Best in: Claude.

Here are [number] customer comments and reviews about a competitor: [paste]. Cluster them into themes with a count for each theme. For the five largest themes, quote the two most representative original comments. Separate product complaints, service complaints, and unmet expectations clearly. Do not infer overall brand sentiment from this sample; report the counts as they are.

18. Negative reviews as an opportunity map

Best in: Claude or ChatGPT.

From this set of negative competitor reviews [paste], list every problem that appears at least three times. For each problem answer three things: is it a product, process, or expectation problem; is [our brand name] genuinely better on that point given our product description here [paste description] — answer honestly, including when the answer is no; and if yes, how we could prove it on our product page without naming the competitor.

19. Their creator selection pattern

Best in: ChatGPT.

Here are the creators working with the competitor [name]: [paste handles, audience sizes, and content types]. Analyze their selection pattern: audience size range, niche types, content formats requested, and how often collaborations repeat with the same person. Then propose creator selection criteria for [our brand name] that avoid audience overlap with them, and state the trade-off each criterion carries.

20. Seasonal campaign read

Best in: ChatGPT or Gemini.

Here is [seasonal moment] campaign material from [3 competitors]: [paste]. Compare when they started, the offers used, the story angle, and the channels they prioritized. Draw out the timing pattern in this category, then build two campaign calendar options for [our brand name]: one that follows the market pattern, one that deliberately moves earlier. Name the specific risk in each.

Prompts 21-25: How do you turn findings into decisions?

This is the block most teams skip, and it's where competitor analysis actually pays. All five take the output of earlier prompts as their input.

21. A SWOT that isn't generic

Best in: Claude.

Using all of the findings below [paste summaries of the earlier analyses], build a SWOT for [our brand name] under three strict rules: every point must name a competitor or specific evidence, no point may be equally true of every company in the category, and every weakness must be followed by one action that could start within 30 days. Delete any point that fails those rules, and tell me how many you deleted.

22. Sales battle card

Best in: Claude.

Write a one-page battle card for [our brand name]'s sales team when they meet [competitor name]. Include: why customers choose them, why customers leave them, three questions our reps should ask to surface their weaknesses, the three most common objections with short answers, and one honest differentiating sentence. Do not use any claim we cannot back with our own product or data.

23. Attack simulation from their side

Best in: Claude — the output is sharpest and least comfortable.

Play the role of head of growth at [competitor name]. You have just studied [our brand name]: [paste our product description, positioning, and homepage copy]. Write an internal memo to your CEO answering three things: where they are most vulnerable, what you would do in the next 90 days to take their share, and which part of their position you would genuinely struggle to copy. Write it sharp and honest, not polite.

24. Opportunity prioritization

Best in: ChatGPT or Claude.

Here are the opportunities that came out of our competitor analysis: [paste list]. Score each on three axes: potential impact on [our core metric], effort required in person-weeks, and how quickly a competitor could copy it. Present it as a table sorted by impact divided by effort, then recommend three to run this quarter and state what we would have to drop to do them.

25. Monthly change report

Best in: Claude.

Here is last month's competitor snapshot: [paste]. Here is this month's: [paste]. List only what changed across five things: pricing, messaging, features, channels, and content frequency. For each change, give one sentence of interpretation and tag it with one of three labels: ignore, monitor, or respond now. Do not repeat anything that stayed the same, and do not add inferences that aren't visible in either snapshot.

How much of the AI's output do you need to verify by hand?

Verify every number, date, and product name — the rest, meaning structure and interpretation, is usually usable after one careful read.

The part we cut from first drafts is always the same kind of thing: competitor plan pricing that changed months ago, invented follower counts, and launch dates off by a year. That's why nearly every prompt above carries a version of the same line — make the model flag what it can't verify. That line doesn't remove errors; it moves them somewhere you can see them, which is the difference between a report you can use and one you have to re-check end to end.

One other habit earns its keep: paste the raw material. Since we stopped asking "how is competitor X's content" and started pasting the article itself, the number of claims we had to correct dropped sharply — not because the models changed, but because there was no longer room to improvise.

What these prompts can't do

• They're not a data source — no model has access to a competitor's traffic, revenue, ad spend, or internal numbers. If a figure like that appears, it was invented, not found.
• They don't replace customer research — competitor analysis explains what rivals do, not what your buyers actually need.
• Snapshot, not monitoring — output from prompts 1 to 24 expires the moment a competitor edits their pricing page. Only prompt 25 is built to repeat.
• Quality follows the input — paste five reviews and you get a five-review analysis, however tidy the output looks.

Which of the 25 should you run this week?

Run three: prompt 3 to surface rivals missing from your list, prompt 5 to see which brands AI names in your category, and prompt 23 to read your own position from the other side of the table. All three fit in one afternoon, and they usually produce one surprise that reorders the quarter.

Worth stating plainly: Intura sells AI Search Optimization services, so weigh the recommendations in prompts 5 and 9 with that in mind. Both still run fine manually with three free accounts and a spreadsheet — the only thing that changes is how much of your week they take.

If you want to see which brands ChatGPT, Claude, Gemini, and Google AI Overview name in your category without retyping prompt 5 every week, that's the part we build at Intura. Start with the prompts; reach for tooling only once the frequency starts to hurt.