The report button is the last step, not the first, and the gap between operators who get fake reviews removed and operators who do not is almost entirely about what exists before the button gets pressed.
A report that says this review is fake, we have never served this person is a claim. A report that says this review describes a Saturday booking under this name at 19:30; we have no booking, no cover, and no card transaction matching it; the same account has posted eleven one-star reviews of restaurants in four counties in nine days is a file.
The second one gets actioned. Not always, but at a completely different rate. Building it takes about twenty-five minutes and almost none of that is writing.
Step one, within the hour: capture
Reviews get edited and accounts get deleted. If either happens before you have captured the original, you have nothing — including nothing to show a franchisor, an insurer or a solicitor if this ever becomes serious.
Capture all of this:
- Full-screen screenshot showing the review, the reviewer's display name, the star rating, the stated age of the review, the page URL and your device clock. Not a crop of the text.
- The reviewer's profile page, if it is public: their other reviews, review count, and the dates. This is usually the single most persuasive item in the file and it is the one that disappears fastest.
- The review text as plain text, pasted into your log. Searchable text lets you spot the same wording appearing on another listing later, which a screenshot does not.
- A second capture from a signed-out device to establish it is publicly visible rather than something only you can see.
Do this before you tell anyone, reply to anything, or discuss it with the team. It is the only step with a deadline.
Step two: find the policy breach and name it
This is where most reports die. A review being untrue is not, on its own, a ground for removal — platforms do not adjudicate whose account of a Saturday night is correct. What gets acted on is a content policy breach, and your report needs to name one.
The ones that actually apply to restaurants:
Not a genuine experience. No visit took place. This is the strongest ground available and the one that requires your records rather than your opinion.
Conflict of interest. Posted by a competitor, a former employee, or someone in a dispute with you. Needs supporting evidence — an employment date, a matching name, a prior complaint.
Off-topic. About a different business, a delivery courier, a landlord dispute, or a political opinion. Common and frequently successful, because it is verifiable from the text alone.
Harassment or personal information. Names a staff member and makes personal allegations, or publishes someone's details. Treat this as urgent and separate; it has a duty-of-care dimension as well as a reputational one.
Impersonation or clearly inauthentic account. A profile created the same day, no photo, a string of reviews across unrelated businesses in distant places.
Pick one. Reporting a single review under four categories at once reads as a scattergun and weakens all of them.
Step three: pull the operational records
This is the part that distinguishes a file from a grievance, and it is specific to hospitality — you have more disprovable detail available than most businesses.
If the review names a date, a time, a dish, a staff member or a party size, check each against:
| The review claims | Check against | What it establishes |
|---|---|---|
| A date and time | Booking system, cover count for that service | Whether a party of that size was in at all |
| A named dish | Menu on that date, 86 list | Whether you were serving it |
| A specific bill amount | Till records for the service | Whether a transaction near that value exists |
| A named staff member | Rota for that shift | Whether that person was working |
| A long wait | Ticket times for that service | Whether the service ran as described |
| Card details or a refund | Payment records | Whether any transaction exists |
Three of these contradicting each other is a strong file. One is suggestive. None, and you may well be looking at a real guest whose memory of a bad evening has drifted — which happens constantly and is not fraud.
Two things worth keeping straight. Absence of a booking is weak on its own if you take walk-ins; say so in the report rather than letting it be found. And a dish not on the menu that day is unusually strong, because it is checkable, specific, and not something a real guest gets wrong.
Step four: write the report in five lines
Factual, dated, one policy ground, no adjectives.
Review posted 14 September by "[display name]", 1 star, describing a meal on Saturday 6 September at approximately 20:00 for six people, including the lamb shank.
We have no booking for six on that date and no till transaction above £90 during that service. The lamb shank came off the menu on 22 August and was not served on 6 September.
The same account has posted nine one-star reviews of restaurants across three counties since 5 September, all of comparable length.
Reporting under: not a genuine experience.
Captures and records attached.
What makes that work: a specific falsifiable claim, the records that contradict it, a pattern, one named ground. What is absent: any mention of being unfair, damaging, or devastating to a small business. Those are true and they do not help.
Step five: the log you keep regardless of outcome
One appended file, four columns.
| Date | Review and account | Ground reported | Outcome and date |
|---|
It earns its keep in three situations. A repeat from the same account — the second report citing the first is far stronger. A pattern across several of your sites, which is a different and more serious conversation. And anyone later asking what you did about it, where a dated record is the difference between diligence and a story.
What to do while you wait
Decide on a reply separately from the report. Reporting and replying are independent. A short, non-committal, dated reply often serves you better than silence, because the people reading are not the people adjudicating. Do not say the review is fake in public — if it is removed you have an odd orphaned reply, and if it is not you have an argument on your profile.
Do not chase it with favourable reviews. A sudden burst of five-star reviews is the pattern most likely to trip filtering, and it costs you the genuine ones alongside the engineered ones. What actually crosses the line when you ask is worth knowing exactly.
Do not report the same review repeatedly. It does not escalate anything.
Expect nothing most of the time. A well-built report on a clear policy breach has a decent chance. A well-built report on a review that is merely unfair has almost none, and the file was still worth building — because in six months, if this turns out to be one of several, the file is what makes the pattern arguable.
When there is no case to build
Reviews with no detail are effectively unreportable. A one-star with no text gives you nothing to contradict. Frustrating, and there is no route.
Your records have to be good enough to rely on. If covers are not counted per service and the rota lives in a photo on somebody's phone, step three is unavailable. That is worth fixing for its own sake.
A real guest can be wrong about everything. Wrong date, misremembered dish, wrong staff name. Before concluding a review is fabricated, check the surrounding week — a review describing your Saturday that actually happened on the Friday is the single most common false positive here.
Reviews that vanish are a different problem entirely. Missing reviews are usually not removal, and panicking about it costs real time — telling removal apart from a dry spell comes first. And if several arrived at once rather than one, a burst needs a different read altogether.
OMMU captures review text and reviewer history as reviews land, which means the evidence exists before you know you need it — the part of this that cannot be done retrospectively. A shared folder and a disciplined screenshot habit achieves the same thing, as long as somebody does it within the hour.