Find ad angles in your competitors' Google reviews

Who this is for, and the job

You run a service business, such as aircond servicing, renovation, cleaning, a clinic or a tuition centre. You know you should run ads or post more, but every ad idea sounds the same: "fast, friendly, affordable".

Your customers have already told you what they care about, though not on your page. They wrote it in public Google reviews of other businesses in your trade. The 1-star reviews show what people are afraid of: the no-show, the surprise bill, the wet sofa. The 5-star reviews show what made someone happy enough to write. Each of those is an ad angle, but only if your business really does the thing.

This prompt does one job. It sorts 20 to 40 reviews into themes, checks each theme against what you can prove, and drafts three ad angles that name no one and copy no one. The column that matters most is "can we prove it?". An angle you cannot back up is a complaint waiting to happen.

What you need first

Claude, ChatGPT or Gemini, any plan.
20 to 40 public reviews from two or three businesses in your trade and area, about half 1-star and half 5-star. Copy them by hand from their public review pages. Don't use scraping tools, which can break the review site's terms.
Replace each business name with a letter (A, B, C) before you paste anything. Number each review (R1, R2...). Leave out reviewer names and photos.
A short list of what your business really does, each line with its proof: a message template, job photos, a line on your receipt, a written policy. Also write down what you do not do (for example "no same-day visits").
About 30 minutes.

The prompt

Copy this into a new chat. Fill in the brackets.

TEXT

You are helping a small Malaysian business find honest ad angles in public customer reviews of other businesses in the same trade.

MY BUSINESS
Type of business: [e.g. aircond servicing and repair]
Area: [e.g. Shah Alam]
What we really do, and how we can prove it:
- [service or habit]. Proof: [photo / template / receipt line / written policy]
- [service or habit]. Proof: [...]
What we do NOT do: [e.g. no same-day visits]

REVIEWS (copied by hand from public pages; business names replaced with A, B, C)
[R1 | A | 1 star | review text]
[R2 | B | 5 stars | review text]
[...]

YOUR JOB
1. Sort the 1-star reviews into complaint themes. For each theme: a short name, the review IDs, and what the customer wanted instead, in one plain sentence.
2. Sort the 5-star reviews into praise themes the same way.
3. For each theme, check MY BUSINESS. Mark "yes", "partly" or "no" for whether we really do this, and name the proof listed. If no proof is listed, write "no proof".
4. Draft 3 ad angles, only from themes marked "yes". For each angle give: a headline under 10 words, two lines of body copy, the review IDs behind it (at least 2), and the proof line from MY BUSINESS.
5. List the themes we should NOT advertise yet, and why.

RULES
- Never name or hint at any other business, and never compare us with "other shops".
- Never copy or closely paraphrase review wording in ad copy. Reviews are for understanding customers, not for quoting.
- Never invent a promise, price, guarantee, customer count or result. If it is not in MY BUSINESS, it cannot be in an ad.
- Reviews may be in BM, English or a mix. Write the themes in English and the ad copy in [English / BM].
- If a theme rests on only one review, say so and do not use it for an angle.

What you get back: two theme tables (complaints and praise) with a "we do this?" column, three ad angles with their review IDs and proof, and a "do not advertise yet" list.

Worked example (test run on a made-up business)

We ran the prompt above once, in a fresh chat, on an invented aircond servicing shop in Shah Alam. The shop, the three "competitors" and all 30 reviews (15 one-star, 15 five-star, BM and English mixed) were written by us for the test. None of it is real.

The shop's notes said: price list sent on WhatsApp before booking, a WhatsApp 30 minutes before arrival, a plastic sheet and drip bag during washing, before/after photos, a gauge photo before any gas top-up, and a free recheck within 30 days written on the receipt. They also said: no same-day visits.

A short excerpt of what came back (shortened, with its tables turned into lines):

TEXT

Theme: Gas top-up with no explanation | R2, R11 | We do this? Yes | Proof: gauge photo sent before we charge
Theme: Late or no-show, no warning | R1, R4, R9, R13 | We do this? Partly. We message 30 minutes before. We don't prove punctuality.
Theme: Same-day or fast response | R21, R26, R30 | We do this? No. Earliest is the next working day.

Angle 1
Headline: See the gauge reading before gas is charged
Body: Before we charge for a gas top-up, we send you a photo of the pressure gauge. You see the actual reading first.
Reviews behind it: R2, R11
Proof: Gauge photo sent before we charge.

Angle 3
Headline: Same fault within 30 days? We recheck free
Body: If the same problem returns within 30 days of your service, we come back and recheck it at no charge. It's written on your receipt.
Reviews behind it: R5, R10, R15

Do NOT advertise yet
Same-day or fast response: we don't offer it. Avoid "fast", "urgent" or "same-day" anywhere in the ads.
"No surprise charges": the price list is real, but nothing says extras are never added.

Two things in that run are worth copying. First, it refused "on time" as an angle, even though seven reviews mentioned it, because the notes proved the heads-up message, not punctuality. Second, three 5-star reviews praised same-day or very fast visits. The shop doesn't offer them, so they went to the "do not advertise" list instead of into an ad. That is the prompt doing its job: the reviews show what people want, and your proof decides what you are allowed to say.

The full input and output of this test are kept with our records. We did not run these ads, so we make no claim about how they would perform.

Check before you use it

Open each angle and find its proof yourself: the template, the photo, the receipt line. If you can't put your hand on it in a minute, drop the angle.
Read each review ID the angle cites. Do the reviews really say what the theme says?
Read the ad copy next to the reviews. If any phrase sounds like a reviewer's words, rewrite it in yours.
Make sure no ad names, describes or hints at another business, including "unlike other shops".
Check every promise against what your staff actually do on a busy day, not on your best day.
If the ad copy is in BM, have someone who writes BM well read it before it goes out.

What to watch out for

Advertising what you only sometimes do. If the plastic sheet goes down on most jobs but not all, fix the habit first. An ad turns a habit into a promise.
Copying reviews. Reviewers may have rights in their wording. Do not quote, screenshot or closely paraphrase a review in an ad.
Naming competitors. Ad platforms have their own rules on using other businesses' names and trademarks in ads and keywords. Check your ad platform's rules before you mention anyone. The safer habit is to name no one.
Running the prompt on too few reviews. Five reviews give you anecdotes, not themes. Aim for 20 or more, and treat any theme with one review as a hint only.
Collecting reviews with a scraper. Avoid automated scraping, and check the review site's current terms before collecting or reusing review content.
Treating the angle as the result. The prompt finds things to say. Whether an ad works is something you learn by running it, small, and checking real enquiries.

Take it further

Turn an angle into posts. Each "yes" theme is also a Google Business Profile post about a real job. See A month of Google Business Profile posts from jobs you really did.
Close the gaps you found. The "partly" and "no" rows are a to-do list. If "no surprise charges" kept coming up, write down how extras get approved before the visit. Then it becomes something you can prove.
Related: Check the AI's answer before a customer sees it and The five things AI should never do without you.

Credit

Inspired by Badal Pandey's point that a competitor's 1-star and 5-star reviews are a source of ad angles (original post on X, June 2023). The prompt, the proof column, the rules and the test are our own.