RestaurantsSEO.com
Guide 03 · Restaurant SEO › Reviews

Restaurant reviews: the thresholds moved, and most restaurants have not noticed

In one year the share of diners who will only use a business rated 4.5 or higher nearly doubled, from 17% to 31%. Meanwhile 74% only weigh reviews from the last three months, which means a brilliant rating built on old reviews now reads as a restaurant that used to be good. This guide covers the arithmetic, how to ask without breaking federal rules, and what to write back.

Updated 12 August 2026 ~27 min read Includes FTC compliance
Twelve-month projection
Reviews in 12 months
Recency window
Rating in 12 months
46
1
4.2

Two things move together and most operators only watch one. Volume builds slowly and is easy to see. Recency expires quietly — if you are collecting fewer than roughly four a month, your newest review keeps drifting outside the ninety-day window that three quarters of diners actually weigh. The rating projection assumes new reviews average 4.7, which is typical for a restaurant that asks its happy guests rather than waiting for volunteers.

The standards

What diners now require before they will consider you

These are consumer-behavior thresholds, not ranking factors. Each one is a point at which a measurable share of people stop considering your restaurant — before they read a word about the food.

Restaurant review thresholds and what changed year over year
ThresholdWhat it meansMovement
4.0 stars68% of consumers require at least a four-star rating before they will consider a business at all.from 55%
4.5 stars31% will only use a business rated 4.5 or higher. This is now the working target, not a stretch goal.from 17%
20 reviews47% will not use a business with fewer than twenty. Only 9% will use one with five or fewer.volume floor
90 days74% only weigh reviews written in the last three months.recency window
14 days32% specifically look for reviews from the last two weeks.from 20%
Every reply80% are more likely to use a business that answers all its reviews. 42% are unlikely to use one that never replies.response rate
Same day19% now expect a same-day response; 81% expect one within a week.from 6%
97% / 41%Consumers who read reviews for local businesses, and the share who now “always” read them before choosing.from 29%
50%Consumers put off by generic or templated review responses.reply quality
83%Of people who were asked for a review, the share who actually wrote one.the opportunity

How many Google reviews does a restaurant need?

At least twenty, with fifty or more as a working target. 47% of consumers will not use a business with fewer than twenty reviews, and only 9% will use one with five or fewer. But volume alone is not the goal — recency matters as much. Because 74% only weigh reviews from the last three months, you need roughly four new reviews a month, every month, to keep your newest review inside that window all year. Not a campaign. A habit.

The reassuring number in that table is the last one. 83% of people who were asked for a review wrote one, and the share who say they will “always” write one when asked rose from 16% to 28% in a year.

The bottleneck is not consumer reluctance. It is that most restaurants never ask, ask once and stop, or ask in a way that is easy to ignore. That is a solvable operational problem rather than a marketing one.

A parent and child sharing pizza at a restaurant table
The people most likely to write you a good review are sitting in your dining room right now, and most of them will never be asked.

Volume is visible and builds slowly. Recency expires quietly. Almost every restaurant that plateaus is losing on the second one.

The trap

The recency problem that catches good restaurants

This is the failure mode that specifically punishes restaurants that were doing well. It is worth understanding because the symptom looks like nothing is wrong.

The pattern is consistent. You open. You are excellent. In eighteen months you collect two hundred reviews and settle at a 4.7. Then you get busy running a restaurant, and asking for reviews stops being anybody's job.

Three years later the number on your profile still says 4.7. Nothing looks broken. But your newest review is eight months old.

To a diner scanning quickly, that reads as a restaurant that used to be good — and given that 74% only weigh reviews from the last three months, most of what you earned has quietly stopped counting in their heads. To Google and to AI systems, stale reviews are a weak freshness signal, and freshness has been weighted more heavily since 2026.

The rating you worked for stops working for you, without ever going down.

Why four a month

The arithmetic is simple. The ninety-day window is about three months. To always have a review inside it, you need reviews arriving at least monthly, and you need enough margin that a slow month does not empty the window entirely.

Four a month gives you roughly twelve inside the window at any time, which reads as an active restaurant, survives a bad week, and compounds to nearly fifty a year. For most independents that is one review per two hundred covers or so — a genuinely modest ask.

The projector at the top of this page shows what happens at your current rate. If you are below four, the window is where you are losing, regardless of how good your rating is.

The other half: rating math

Ratings are averages, which means they get harder to move as volume grows. This is worth internalizing before you set a goal.

A restaurant at 4.2 with 40 reviews needs roughly 25 more at five stars to reach 4.5. The same restaurant at 4.2 with 300 reviews needs closer to 190.

Two consequences. Fix the underlying problem first, because adding volume to a broken experience just locks in the average. And if you are early, protect the rating aggressively — the reviews you collect in your first year set a floor that is expensive to raise later.

A habit that survives a busy Friday

  • One person owns it. Not everyone, which means nobody. A manager who asks two tables a shift.
  • Ask at the moment of genuine satisfaction — when a guest says something kind, not when the check lands.
  • Make it one tap. A short link or QR on the check presenter, replaced when it looks worn.
  • Check reviews at open and at close. Ten minutes twice a day beats a quarterly session.
  • Write the monthly count on the same board as covers, so it is visible when it slips.
The system

How to actually get reviews, without any of the above

Everything that works here is unglamorous and free. The constraint is consistency, which is why the goal is a routine rather than a push.

01

Ask in person, at the right moment

The single most effective method and the one restaurants use least. When a guest says something genuinely kind — about a dish, about the service, about the room — that is the moment. “That means a lot. If you have thirty seconds, a Google review would help us enormously.”

Not at the check. Not to every table. From the person they actually connected with.

02

Make it one tap

A short link or QR on the check presenter, a card, or a table tent. Google provides a direct review link from your profile dashboard — use that rather than sending people to search for you, because every extra step loses a share of them.

Replace the card when it starts looking worn. A greasy, faded QR reads as neglect.

03

Follow up if you have a channel

If you take reservations, run a waitlist, or have an email list, a short message the following day works well. Keep it to two sentences, link directly, and send it to everyone rather than filtering by predicted sentiment.

Once per visit. Restaurants that chase reviews with repeated reminders train people to ignore them.

04

Spread across platforms deliberately

The average diner consults about six sources. Google first, but Yelp, Apple Maps, Tripadvisor and OpenTable all matter, and Yelp content is now licensed into ChatGPT, which quietly makes it part of your AI presence.

A workable approach: default everyone to Google, and when a guest mentions they found you somewhere else, ask for that platform instead.

05

Give staff a reason to care

Reviews that name a server by name are the most persuasive kind, and they are also the most motivating for the person named. Read them out at pre-shift. Post them in the back.

Do not pay per review. It creates exactly the pressure that produces gating, fake accounts and staff soliciting from friends.

06

Fix the thing people complain about

If three reviews in a month mention the wait, no amount of asking will outrun it — you will just collect more reviews mentioning the wait. Reviews are a measurement instrument before they are a marketing channel.

The restaurants with the best ratings are usually not the best at asking. They are the ones with fewer things to complain about.

How do I get more Google reviews for my restaurant?

Ask, consistently, in person, at the moment a guest expresses genuine satisfaction — and make the act of writing one a single tap via a direct review link on the check presenter. 83% of people who are asked write a review, so the constraint is almost always that nobody is asking rather than that guests are unwilling. Give one manager per shift ownership of asking two tables, track the monthly count somewhere visible, and aim for roughly four a month to stay inside the ninety-day recency window. Do not incentivize by sentiment, do not filter who gets asked, and do not let staff write them.

The reply

Responding: write for the audience, not the reviewer

The person who wrote a one-star review has largely made up their mind. The fifty people who read your reply have not. That single reframe fixes most bad restaurant replies.

The numbers make the case for replying at all: 80% of consumers are more likely to use a business that answers all its reviews, 42% are unlikely to use one that never replies, and 19% now expect a same-day response — up from 6% a year earlier.

Partial response measurably underperforms full response. Answering only the positive ones, or only the negative ones, both score worse than answering everything. And templated replies actively hurt: 50% of consumers are put off by generic responses, so a canned paragraph under a detailed complaint reads as indifference.

One 2026 change worth knowing: review replies now pass through a moderation queue before publishing rather than appearing instantly. It usually clears quickly, so do not assume a reply failed to post.

The shape of a good reply to a bad review

  • Acknowledge the specific thing. Not “we're sorry you had a bad experience” but “a forty-minute wait on a Tuesday is not acceptable.”
  • Say what you are doing about it, if there is anything true to say. Vague promises to “do better” read as filler.
  • Offer to continue off the platform. A name and an email address, not a phone tree.
  • Stop there. Short replies read as confident. Long ones read as defensive.

And what not to do

  • Argue about the facts, even when you are right and can prove it
  • Mention what the guest ordered, when they came, or anything that identifies them
  • Imply they are lying, or that they are a competitor
  • Copy the same apology under fifteen different complaints
  • Reply while angry — write it, leave it an hour, then read it as a stranger would
★☆☆☆☆

“Waited 45 minutes for a table we had reserved, then another 30 for food. The tacos were good but we missed our movie. Won't be back.”

What not to write

“We're sorry to hear you had a less than perfect experience. We strive to provide excellent service to all our guests and we hope you'll give us another chance. Please contact us at your convenience.”

Addresses nothing specific, could be pasted under any complaint, and asks the guest to do the work of following up. This is the reply 50% of readers say puts them off.

What to write

“A 45-minute wait on a reservation is a failure on our part, and adding a 30-minute ticket time to it made a bad night worse. We overbooked the 7pm block that Saturday and have cut the number of reservations we take in that window. I'd genuinely like to make it right — I'm Maria, the GM, at maria@casaverde.com. Glad the tacos held up, at least.”

Names the specific failures, states a concrete change, gives a real person to contact, and stays short. The reader learns the restaurant noticed and fixed something.

How quickly should a restaurant respond to reviews?

Within the week at minimum — 81% of consumers expect a reply within seven days — and same-day is now expected by 19%, up sharply from 6% a year earlier. The practical version for a restaurant is ten minutes at open and ten at close, which handles nearly all volume without becoming a project. Batching twenty replies onto one afternoon every few months is visible to anyone reading the profile and reads exactly as what it is.

Damage control

Reviews you believe are unfair or fake

Every restaurant eventually gets one: a review from someone who never came in, a competitor, a former employee, or a guest describing an event that did not happen. Here is what can actually be done.

What can be removed

Platforms remove reviews that violate their content policies, not reviews that are merely negative or unfair. The categories that genuinely get removed are reasonably narrow:

  • Clearly not a customer — describing a business that is not yours, a location you do not have, or an event you can show did not occur
  • Conflict of interest — competitors, former staff, anyone with a demonstrable stake
  • Off-topic content — political statements, complaints about a neighboring business, rants unrelated to the dining experience
  • Prohibited content — harassment, slurs, personal information, threats
  • Spam or duplicates — the same review posted repeatedly, or obvious bot activity

What does not get removed: a genuine customer describing a genuinely bad night, however unrepresentative you feel it is, and however much detail they got wrong.

How to flag one properly

Flag it through the platform with a specific, factual reason tied to the policy it violates. “This is unfair” will not work. “This review describes a party of twelve on a date we were closed for a private event” might.

Reply publicly while you wait, calmly and briefly, because the reply is what future readers see regardless of the outcome. Something like: “We have no record of this visit and have asked Google to review it. If we have got that wrong, please contact me directly — I would like to understand what happened.”

What not to do: threaten legal action to force removal. Beyond rarely working, review suppression through unfounded legal threats is specifically addressed by the FTC rule.

The proportionate response

A single one-star review among fifty is statistically invisible and functionally harmless. The reply matters more than the review does.

What genuinely damages a restaurant is a pattern — five recent one-stars all mentioning the same thing. That is not a reputation problem to be managed, it is an operations problem showing up in public.

The best defense against an unfair review is volume of fair ones. A restaurant collecting four a month absorbs a bad one without anyone noticing. A restaurant that has not collected any since spring does not.

Two 2026 changes

Anonymous reviews. Platforms now permit reviews under pseudonyms in some cases, which raises volume and makes it harder to match a review to your own records. Reply to the substance rather than trying to work out who it was.

Reply moderation. Replies pass through a queue before publishing. If yours has not appeared, give it time before rewriting it.

A laptop showing analytics charts on a desk
Watch the pattern, not the individual review. A dashboard is useful for spotting the former; the latter is best handled by reading it and replying like a person.
The new stakes

Your reviews are now the raw material for AI descriptions of you

This is the development most operators have not yet absorbed. 82% of consumers read AI-generated review summaries — the automatic synopses that now appear above reviews on major platforms — and 23% say they would be happy to decide on that summary alone.

Those summaries are generated from your review text. Which means the sentences your guests write are being synthesized into the description a machine gives of your restaurant, and shown to people who will never scroll down to read the reviews themselves.

Google went further and replaced its Business Profile Q&A section with Ask Maps, where Gemini answers visitor questions automatically using your profile, your website and your reviews. You can no longer pre-write the answer to “is it loud in there?” — but your reviews are one of the three sources that answer gets built from.

What follows from that

  • Recent reviews carry disproportionate weight, because summaries lean on current sentiment. A themed complaint from eighteen months ago that you have since fixed can persist in a summary if nothing newer contradicts it.
  • Specific praise is more useful than generic praise. “Great place” gives a summary nothing. “The birria is the best I've had in Denver and the patio is quiet” supplies attributes that get matched against real queries.
  • Fixing an operational problem now has a compounding payoff — you stop generating the complaint, and newer reviews gradually displace it in what the machine says about you.
  • Your replies are part of the record too, and a pattern of specific, responsive replies reads well to both humans and systems summarizing them.

The full treatment of how assistants use this is in the AI search guide.

Two guests laughing over dinner in a warmly lit restaurant
What a guest writes about this evening becomes, a few weeks later, part of how a machine describes your restaurant to a stranger.

Do reviews affect whether AI recommends my restaurant?

They appear to, in two distinct ways. Reviews are a freshness and prominence signal, so a profile with recent activity looks like an operating business worth recommending. More directly, review text is raw material: AI review summaries are generated from it, and assistants building a recommendation quote attributes that reviewers mention. A restaurant whose recent reviews specifically name dishes, describe the atmosphere, and mention practical details gives an assistant far more to work with than one whose reviews all say “great food, will return.”

Reviews are a system, not a burst

Asking, replying within the week, watching for patterns, keeping four a month coming. It is not hard and it is relentless. We run it for restaurant clients alongside the rest — or call and we will help you set it up to run yourself.

1-800-481-8638
Field notes

Review mistakes we find most often

01

Gating without knowing it

Reputation software that surveys first and shows the Google link only to happy respondents. Sold as a feature, prohibited by platforms, and in the territory the FTC rule addresses. Check what your tool actually does.

02

Asking once, then stopping

A burst of thirty reviews in one month followed by silence looks exactly like what it is, and leaves the recency window empty for the rest of the year.

03

Replying in batches

Twenty replies posted the same afternoon, months after the reviews. Meanwhile 19% of diners now expect same-day. Ten minutes twice a day handles it.

04

The same apology fifteen times

Half of consumers are put off by templated replies. A generic paragraph under a specific complaint communicates that nobody read it.

05

Family and staff reviews at opening

Well-intentioned, easy for platforms to detect from shared devices, and squarely within the undisclosed-insider provision of the FTC rule.

06

Chasing volume over the actual problem

If the reviews all mention the wait, more reviews just means more mentions of the wait. Reviews are a measurement instrument before they are a marketing channel.

Questions

Restaurant review questions

How many Google reviews does a restaurant need?

At least twenty, with fifty or more as a working target. 47% of consumers will not use a business with fewer than twenty reviews, and only 9% will use one with five or fewer. Volume alone is not the goal though — because 74% only weigh reviews from the last three months, you need roughly four new reviews a month, every month, to keep your newest review inside that window all year. A restaurant with 200 old reviews and nothing recent performs worse than one with 60 that arrive steadily.

What star rating does a restaurant need in 2026?

4.0 is the functional minimum and 4.5 is the target. 68% of consumers require at least four stars, up from 55% a year earlier, and 31% will only use a business rated 4.5 or higher, up from 17% — a threshold that nearly doubled in twelve months. Only about 10% insist on a perfect five, so a 4.6 or 4.7 satisfies nearly everyone. The important detail is the speed of that change: a rating that was comfortably acceptable in 2025 may now be disqualifying for almost a third of searchers.

How do I get more Google reviews for my restaurant?

Ask, consistently, in person, at the moment a guest expresses genuine satisfaction, and make writing one a single tap using the direct review link from your Google profile. 83% of people who are asked write a review, so the constraint is almost always that nobody is asking rather than that guests are unwilling. Give one manager per shift ownership of asking two tables, track the monthly count somewhere visible, and aim for roughly four a month. Do not incentivize by sentiment, do not filter who gets asked, and do not let staff write them.

Can I offer a discount for a Google review?

Only under narrow conditions, and platform rules may still prohibit what federal rules permit. Under the FTC Consumer Review Rule an incentive is problematic when conditioned on the reviewer expressing a particular sentiment, so a discount for a five-star review is out. An incentive offered for an honest review regardless of what it says, and disclosed in or near the review, is the permissible version. However Yelp prohibits incentives in its review flow entirely, and Google discourages transactional schemes. Something legal federally can still get your listing flagged, so simply asking is safer and works nearly as well.

Is review gating illegal?

Review gating — surveying guests first and routing only the happy ones to a public review link — is prohibited by major platform policies and sits directly in the territory the FTC Consumer Review Rule addresses, because the resulting public rating misrepresents the actual distribution of customer experience. Enforcement posture suggests a gating flow can count as a violation for each affected consumer, with civil penalties running to $53,088 per violation as of 2026. Plenty of reputation software still sells this as a feature. If your tool shows the Google link only to people who answered positively, that is gating. This is orientation rather than legal advice.

How should I respond to a bad restaurant review?

Write for the audience rather than the reviewer, because the person who wrote it has largely decided and the people reading your reply have not. Acknowledge the specific failure rather than offering a generic apology, state one concrete thing you have changed if there is anything true to say, give a real name and email to continue off the platform, and stop there. Never argue about facts in public even when you can prove you are right, never identify the guest or what they ordered, and never imply they are lying. Short and specific reads as confident; long and defensive does not.

How quickly should a restaurant respond to reviews?

Within the week at minimum, since 81% of consumers expect a reply within seven days, and same-day is now expected by 19% — up sharply from 6% a year earlier. The practical version is ten minutes at open and ten at close, which handles nearly all volume without becoming a project. Batching twenty replies onto one afternoon every few months is visible to anyone reading the profile. Note that replies now pass through a moderation queue before publishing, so a short delay before yours appears is normal.

Can I get a fake or unfair review removed?

Sometimes, if it violates platform policy rather than merely being negative. The categories that genuinely get removed are reviews from people who clearly were not customers, reviews with a conflict of interest such as competitors or former staff, off-topic content, harassment or personal information, and spam or duplicates. A genuine customer describing a genuinely bad night will not be removed however unrepresentative it feels. Flag it with a specific factual reason tied to the policy it breaks, and reply publicly and calmly while you wait. Do not threaten legal action — beyond rarely working, review suppression through unfounded legal threats is addressed by the FTC rule.

Do reviews affect whether AI recommends my restaurant?

They appear to, in two ways. Reviews are a freshness and prominence signal, so recent activity makes a restaurant look like an operating business worth recommending. More directly, review text is raw material: AI review summaries are generated from it, and 82% of consumers now read those summaries with 23% willing to decide on the summary alone. Assistants building a recommendation quote attributes reviewers mention, so a restaurant whose recent reviews name dishes and describe the atmosphere gives an assistant far more to work with than one whose reviews all say great food, will return.

How long does it take to raise a restaurant's star rating?

It depends almost entirely on how many reviews you already have, because a rating is an average and averages get harder to move as volume grows. A restaurant at 4.2 with 40 reviews needs roughly 25 more at five stars to reach 4.5. The same restaurant at 4.2 with 300 reviews needs closer to 190. Two consequences follow: fix whatever is generating the low ratings first, since adding volume to a broken experience just locks in the average, and protect the rating aggressively when you are new because early reviews set a floor that is expensive to raise later.

Should I ask for reviews on Yelp and Apple Maps too?

Yes, though the approach differs. The average diner consults around six sources, and Yelp content is now licensed into ChatGPT, which makes your Yelp listing part of your AI presence whether or not you think about Yelp. Apple Maps usage for finding local businesses nearly doubled to 27% and Apple Business Connect is free to claim. Note that Yelp specifically discourages directly soliciting reviews, so the workable approach is to default everyone to Google and, when a guest mentions they found you on another platform, ask for that one instead.

Are anonymous reviews a problem for restaurants?

They raise volume and make it harder to match a review to your own records, which is genuinely frustrating when you want to understand what happened on a specific night. The practical adjustment is to reply to the substance rather than trying to identify the reviewer — and to resist the temptation to imply in public that an anonymous reviewer is not a real customer, since that reads badly to everyone else reading and is usually wrong. If a review clearly violates policy you can still flag it; anonymity does not exempt it.

Next step

Four a month, answered within the week

That is the whole strategy, and it is free. Everything else on this page is detail around those two numbers.

What defeats restaurants is not difficulty, it is that reviews are nobody's job at eight o'clock on a Friday. If you would rather hand it over — asking systems, response drafting, pattern monitoring, platform coverage — that is something we run. And if you would rather set it up and run it yourself, call anyway and we will tell you how we would do it.

If you are currently using a tool you suspect might be gating, that is worth a call before it becomes a problem.

Related reading