Platform prioritisation
Which review sources engines and assistants actually consult for your category, established from testing rather than assumed from general advice.
How engines read reviews
Engines read three things: aggregate rating, review volume and recency, and the language inside the reviews themselves. The third matters most for answer surfaces, because a model summarising your category quotes what customers actually said rather than the star average you display.
That third signal is the one most reputation programmes ignore. Chasing an average from 4.3 to 4.5 changes a number. Changing what customers consistently mention — that onboarding was fast, that support answered, that a specific feature worked — changes the sentence an assistant produces when someone asks what you are like to work with.
Recency is weighted more heavily than most teams expect. A strong rating built four years ago and untouched since reads as a business that may no longer be operating as it was. A steady trickle of recent reviews signals an active, current provider, and engines treat that as materially more reliable.
Which platforms matter is category-specific and worth establishing rather than assuming. For local businesses it is overwhelmingly Google. For software it is the review platforms buyers and assistants already consult. For professional services it is frequently industry directories nobody outside the sector has heard of.
The one thing not to do is manufacture the signal. Incentivised reviews, gating requests so only satisfied customers are asked, and purchased ratings all violate platform policies and, in many jurisdictions, consumer protection law. They are also detectable, and detection is a considerably worse outcome than a mediocre average.
What the programme covers
The programme covers identifying which platforms shape your category's answers, a compliant collection process that asks every customer rather than a filtered subset, response handling, review markup where you are eligible, and tracking of how engines characterise your business over time.
Compliant collection
A process that asks every customer, not a filtered subset. Gating requests to likely-positive respondents breaches platform policy and consumer protection rules in several markets.
Language guidance
Prompts that encourage specifics rather than generic praise, because the sentences customers write are what a model quotes when describing you.
Response handling
A pattern for replying, particularly to criticism, since responses are indexed too and a well-handled complaint frequently reads better than an uncomplicated compliment.
Review markup
Structured data where you are genuinely eligible, implemented within the current rules rather than the ones that applied before self-serving markup was restricted.
Reputation tracking
How engines and assistants characterise your business when asked, tracked over time — because that sentence is the actual output, not the star average.
How the programme runs
The programme runs continuously: establish which platforms shape your category, fix any collection practice that breaches policy, build a compliant and consistent request process, handle responses, then track how engines describe you rather than only what your average rating says.
- 01
Establish the platforms that matter
Test what engines and assistants cite when asked about providers in your category. The answer is frequently a platform nobody in the business was monitoring.
→ A prioritised platform list with current standing
- 02
Fix collection practice
Remove gating, incentives and anything else that breaches platform policy. This step sometimes reduces your average, and it removes a liability that was compounding.
→ A compliant collection process
- 03
Build consistent volume
Every customer asked, at a sensible moment, with prompts that encourage specifics. Steady recent volume matters more than a burst followed by two silent years.
→ A running request cadence
- 04
Handle responses
Replies to criticism that acknowledge what is true and state what changed. These are indexed and quoted, and they frequently persuade better than the positive reviews do.
→ A response pattern in use
- 05
Track the sentence, not the star
How assistants describe your business when asked, recorded over time. The characterisation is the output that affects buying decisions, not the aggregate rating.
→ A reputation characterisation trend
Compliant collection compared with what gets penalised
The difference is whether every customer gets the same opportunity to review. Asking everyone is compliant; filtering for likely positives is not, regardless of how it is framed internally. Platforms detect the resulting distributions, and consumer protection regulators have pursued the practice.
| Compliant | Penalised | |
|---|---|---|
| Who is asked | Every customer, same process | Only those expected to be positive |
| Incentives | None, or disclosed and unconditional | Rewards contingent on a positive review |
| Timing | Consistent point after delivery | Only after a known good outcome |
| Negative reviews | Answered publicly | Suppressed or routed to a private form |
| Volume pattern | Steady and continuous | Bursts that platforms flag as inauthentic |
Signals you need this now
You need this when assistants describe your business using reviews that are years out of date, when your collection process only asks satisfied customers, when negative reviews sit unanswered, or when nobody knows which platforms engines actually consult about your category.
- Assistants describe you using reviews years out of date
- Your review requests only go to customers expected to be happy
- Negative reviews sit unanswered for months
- Nobody knows which platforms engines consult for your category
- Review volume arrived in bursts rather than steadily
- Competitors are described more specifically than you are
- A previous agency ran an incentivised review campaign
What clients see
Recency improves quickly once collection is consistent, and that alone changes how current a business appears. Changing how engines characterise you takes longer, because it depends on enough recent reviews mentioning the same specifics to shift what a summary reports.
Where this fits
Related AEO services
Common in these sectors
The adjacent disciplines
Deciding between options
Questions about online review signals
Each answer is written to stand alone in 40 to 60 words — the shape an AI Overview or Perplexity citation lifts. Ships with FAQPage schema.
Can we offer a discount for leaving a review?
Not conditional on a positive one, and most platforms prohibit incentives entirely. Beyond the policy risk, incentivised reviews skew toward generic praise, which is exactly the language that carries least weight when an assistant summarises what customers say about you.
Should we ask only happy customers?
No. Gating requests to likely-positive respondents breaches platform policy and, in several markets, consumer protection law. It is also detectable from the resulting distribution. Asking everyone produces a lower average and a considerably more durable position.
How much do reviews affect AI recommendations?
Substantially, and through the language rather than the rating. A model summarising your category quotes what customers wrote. Consistent specific mentions of a strength shape that summary far more than moving an average by two tenths of a point.
What should we do about a bad review?
Answer it publicly, acknowledge what is true, and state what changed. Responses are indexed and quoted. A well-handled complaint frequently reads more persuasively to a prospect than an uncomplicated positive review does.
Which review platforms matter most?
It is category-specific and worth testing rather than assuming. Local businesses are dominated by Google; software buyers and assistants consult a small set of review platforms; professional services frequently depend on industry directories with no general recognition at all.
Can we use review markup on our site?
Only where you are genuinely eligible under current rules, which restricted self-serving markup some years ago. We implement it where it applies and decline where it does not, since invalid markup undermines trust in your structured data generally.
Find out whether this is your constraint.
Thirty minutes with a senior strategist. We pull your live visibility while we talk and tell you plainly whether a online review signals is what you need — or whether your problem sits somewhere else.