AI Competitor Analysis Guide

What a useful competitor analysis produces

A competitor analysis is useful only if it changes something you do: an offer, a headline, a price presentation, a channel, a page. Most analyses end as a long document nobody opens again. This method ends with three specific decisions and the evidence behind each.

AI makes the collection and sorting fast. It does not make the judgement for you, and it will invent competitor facts if you let it answer from memory. The method below keeps it working on evidence you have gathered.

For a comparison of the tools themselves, see AI competitor analysis tools.

Step 1: define the customer's decision

Write one sentence describing the choice your customer makes. For example: "A practice manager choosing a payroll provider for a 20 person dental practice." Competitors are whoever that person seriously considers, which may include doing nothing, a spreadsheet or an accountant, not just companies that look like you.

Step 2: choose three to five competitors from that decision

Pick from three sources:

• • Who appears on page one of Google for the searches your customers make (use your Search Console queries, not guesses).
• • Who customers mention when they tell you what else they considered.
• • Who is advertising against your core terms (check the Google Ads Transparency Center and Meta Ad Library).

Three well chosen competitors beat ten loosely chosen ones.

Step 3: collect public evidence into one document

For each competitor, copy into a single document:

1. • Their homepage headline and subheadline, and their main offer page.
2. • Pricing or how they present price (even "contact us" is evidence).
3. • The ads they are running in the Meta Ad Library and the Google Ads Transparency Center, with start dates.
4. • Their three most visible proof points (reviews, case studies, logos, figures).
5. • A sample of public reviews that mention what customers liked and disliked.

Paste text, not screenshots, where you can. Label every item with the competitor name and the source URL. This document is the only thing the AI should work from.

Step 4: use AI to extract patterns, with quotes

Give the document to an AI assistant with prompts that force it to cite. Three prompts that work:

• • "For each competitor, list the main promise, the main proof and the main call to action. Quote the exact source text for each."
• • "List every customer objection these ads and pages try to answer. Group them and say which competitor addresses each, with a quote."
• • "What do customers praise and complain about in these reviews? Give counts and example quotes."

Then one comparative prompt: "Based only on this document, where is there a promise, proof type or objection that none of them address well?" The phrase "based only on this document" matters; without it the model fills gaps from general knowledge.

Step 5: verify before you believe

Spot check at least five quotes against the source. Check any number the AI reports (review counts, prices) yourself. Where the AI draws a conclusion ("competitor B is targeting enterprise"), ask what evidence supports it and decide whether you agree. Anything you cannot trace to a source gets deleted.

Step 6: write a one page brief

One page competitor brief template

Section	What goes in it	Example (illustrative)
Customer decision	The one sentence from step 1	Practice manager choosing payroll for a 20 person dental practice
Who they compare	Three to five competitors and why each is included	Two national providers, one local bureau, the practice's accountant
What everyone promises	The shared claims, which are no longer differentiators	Everyone says easy, compliant and supported
What each owns	One distinct position per competitor, with a quote	Provider A leads on price; Provider B leads on dental specific features
Unanswered objection	The concern customers raise that nobody addresses well	Reviews complain about switching mid tax year; no page explains it
Our three decisions	Specific changes, each linked to evidence above	New page on switching mid year; add dental proof to homepage; test a switching support message only if we can deliver it

Step 7: commit to three changes and measure them

Pick three changes you can ship within a month. Attach one measure to each: a page's conversion rate, an ad's click through rate, the share of sales calls where a given objection comes up. Put a date in the diary to rerun steps three to five in a quarter and see what moved, both for you and for them.

Worked example, step by step

Illustrative example: a hypothetical garden design studio sells design packages to homeowners.

• • The customer decision: "A homeowner with a medium sized garden choosing someone to design it before spending on landscaping."
• • Competitors chosen: two local studios that appear for "garden designer" searches in the area, and a national online design service running Meta ads.
• • Evidence collected: three homepages, the national service's 14 active ads (with start dates), and 60 public reviews across the three.
• • AI extraction shows all three promise "a garden you love", the national service leads on low fixed price and speed, and reviews for the local studios repeatedly mention uncertainty about what the build will cost.
• • Verification: the review pattern holds when checked by hand; one AI claim that a competitor "offers free consultations" had no source and was removed.
• • Three decisions: publish a page explaining how build costs are estimated at design stage; add an example budget range to the package page with clear caveats; test an ad angle about knowing the build cost before you commit.

None of those decisions came from the AI. The AI made it possible to read 60 reviews and 14 ads in minutes instead of hours.

Common traps

Copying the leader. If everyone already makes a promise, making it too adds nothing. Look for what is missing.

Analysing only direct lookalikes. The real alternative is often a different kind of solution, or doing nothing.

Trusting AI memory. Models are not current and do not know small competitors. Evidence first, always.

If you would like a second opinion on your competitive position, the free written breakdown from AfM is based on public information and any figures you choose to share, and ends with what we would prioritise.

FAQ

Can I ask ChatGPT or Claude who my competitors are?

You can, but treat the answer as a list of names to check. Models may not know smaller or newer businesses and may describe them wrongly. Your own search results and customer conversations are better sources.

How long does this method take?

For three competitors, collecting evidence takes around an hour and the AI extraction and verification another hour. Writing the brief and choosing decisions depends on how much discussion your team needs.

Should I include competitors' prices in my analysis?

Yes, if they are public, including how they present price. Record the date you saw them, because prices change, and do not copy another business's pricing without understanding their costs.

What if my competitors do not advertise?

Then the ad libraries are quiet and the evidence comes from their websites, reviews and search presence. The absence of advertising is itself useful: it may mean a paid channel is open to you.