Forty-seven percent of consumers will not use a business with fewer than 20 reviews. Only 9% will use one with five or fewer. Those two numbers come from BrightLocal’s 2026 Local Consumer Review Survey, a representative panel of 1,002 US adults, and together they describe something most owners never quite admit to themselves: below a certain count, your rating does not matter at all, because nobody reaches it.
That is the useful shape of the Google reviews statistics for 2026. Not a league table of percentages, but a set of floors you are either above or below. Rating, volume and recency each have one, and failing any single floor removes you from consideration regardless of how well you score on the other two. I want to give those floors names, because naming them is what turns a statistic into a decision.
The Review Trust Floor Has Three Numbers, Not One
Call it the Review Trust Floor: twenty reviews, four stars, ninety days. Those are the three thresholds a local business has to clear before its reputation starts doing any work, and they operate as an AND, never an OR.
Twenty is the volume floor, from the 47% who refuse below it. Four stars is the rating floor, from the 68% who will not go lower. Ninety days is the recency floor, from the 74% who look specifically for reviews written in the last three months. A business with 300 reviews at 4.9 stars where the newest one is from fourteen months ago fails the floor as surely as a business with four reviews from last week.
What makes this framing worth something is that it tells you what to fix first. Most review advice tells you to get more reviews. The floor tells you which one you are failing. If you sit at 4.1 stars with 180 recent reviews, volume is not your problem and asking for more will just dilute slowly. If you sit at 4.9 with eleven reviews, your rating is already fine and you need count. Diagnose, then act. Businesses in the top three local search positions average 47 Google reviews in BrightLocal’s ranking factors study, which is a reasonable target once the floor is cleared, not a starting line.
Volume: Where Diminishing Returns Actually Start
The jump from 5 to 20 reviews moves you from a pool 9% of consumers will consider to one that 53% will consider. The jump from 20 to 47 buys you competitive position rather than basic eligibility. The jump from 47 to 200 buys you very little with human readers, though it still feeds the summarisation layer discussed further down.
This is why review campaigns so often feel like they stop working. They do stop working, at roughly the point where you clear the floor and reach parity with the top three in your category. Teams that keep pushing volume past that point are buying a number that no longer changes behaviour, usually at the cost of the thing that would: recency and response quality.
A practical read on your own position: open the three competitors who rank above you in the local pack, write down their review count and their newest review date, and compare. If their newest review is from this week and yours is from last quarter, you have found your gap, and it is not volume.
Rating: The 4.0 Cliff and the 4.5 Shelf
Ninety-two percent of consumers say star ratings matter when they choose a business, which is unsurprising. The interesting part is where the thresholds cluster. Sixty-eight percent require at least 4 stars. Thirty-one percent require 4.5 or higher. Ten percent say they want 5 stars only.
Read that as a cliff and a shelf. The cliff is at 4.0: fall below and you lose roughly two thirds of the market in a single step. The shelf is at 4.5: climbing from 4.0 to 4.5 recovers a further slice of demand, but it is a slope rather than a drop. And the 10% who demand a perfect 5.0 are mostly unreachable, because a genuine business at scale does not hold 5.0, and consumers who have been online for a decade are suspicious of the ones that do.
The strategic implication is that defending 4.2 is a very different job from climbing to 4.6. Defending means responding fast and resolving the one-star outliers. Climbing means changing something in the service itself, because arithmetic will not get you there: at 150 reviews, moving a 4.2 average to 4.6 takes about 120 consecutive five-star reviews. That is a year of good operations, not a campaign.
Recency Is the Statistic Most Owners Ignore
Here is the shift that has moved fastest. Seventy-four percent of consumers seek out reviews from the last three months. Forty-four percent prioritise the last month. And 32% specifically look at the last two weeks, up from 20% a year earlier.
That last figure is the one I would put on a wall. A twelve-point jump in a single year in the share of people checking the last fortnight means review freshness is compounding in importance faster than volume or rating. It also explains a pattern that confuses owners: a business whose rating and count both look healthy, whose enquiries have quietly declined, and whose newest review is four months old.
Recency also has an operational advantage over the other two floors: it is the cheapest to fix. You cannot add 40 reviews this month without it looking engineered, and you cannot move a rating by half a star at all quickly. You can absolutely ask the last fifteen satisfied customers for a review this week, and that single act clears the ninety-day floor and holds it as long as the asking continues.
Responses Are Read More Than You Think, and Templates Backfire
Ninety-seven percent of consumers who read reviews also read the business responses. BrightLocal’s response research associates responding with up to 18% more revenue, and 80% of consumers say they are more likely to use a business that replies to every review.
Then the catch, which is the part worth internalising: 50% of consumers say generic, templated responses discourage them. Half your audience reads a copied reply as worse than no reply. Expectations on timing have tightened too, with 32% now expecting a response within one day, up from 18% in 2025, and 81% expecting one within a week.
So the response rule is narrow. Reply to everything, reply within a day where you can, and make each reply specific enough that it could not be pasted under a different review. Naming the thing that happened, the person who handled it, or the change you made is what distinguishes a real reply from a form. If you cannot do that at your review volume, reply to the negative ones and the detailed positive ones, and leave the one-line five-stars alone rather than stamping them.
What AI Summaries Did to the Picture
Eighty-two percent of consumers now read AI-generated review summaries. Twenty-three percent rely on those summaries alone to decide. Thirty-nine percent read the summary and then a positive and negative review. Eighteen percent skip summaries entirely. Separately, BrightLocal’s 2026 adoption research put AI tools third among ways people find local businesses, at 45%, up from 6% the year before.
That 23% is the number with the sharpest consequence. For almost a quarter of prospective customers, your reviews are now read by a model and reported secondhand, which means the themes in your reviews matter more than the arithmetic of your average. A summariser reports patterns. Six reviews mentioning slow responses becomes “some customers mention delays,” and it will say that whether your average is 4.2 or 4.7.
Two things follow. First, a cluster of similar complaints is now more expensive than a single furious outlier, which is the reverse of how most owners triage. Second, the specific language customers use in reviews becomes the language the summary uses about you, so prompting for specifics (“if you have a moment, mention which service you booked”) shapes what gets reported. That is not gaming the system. It is giving the system something accurate to work with. We cover the mechanics of that in more depth in our work on AI visibility for local businesses.
Industry Averages Are a Trap
One statistic you will see quoted everywhere is the average star rating by industry, and it is close to useless for planning.
The problem is selection. Published industry averages are computed across businesses that have reviews at all, which excludes the ones that failed early and quietly, and they mix a dentist with eleven reviews into the same mean as a dental group with four thousand. The resulting number tells you almost nothing about what a specific business in a specific city needs to clear. A 4.3 average that looks comfortable against a national benchmark can be last place in a three-mile radius.
Replace the benchmark with a local read. The only comparison that affects your enquiries is the set of businesses that appear alongside you when someone within travelling distance searches your category. Pull that list, record each competitor’s count, average and newest review date, and you have a target that means something. Do it quarterly. It takes about twenty minutes and it is more actionable than any national figure in any roundup, including the ones quoted earlier in this piece.
The same caution applies to review-count benchmarks across categories. A restaurant accumulates reviews far faster than a commercial roofer, because transaction frequency differs by an order of magnitude, so a count that signals health in one category signals an anomaly in another. Compare within your category and within your radius, and treat everything else as background reading.
The Benchmarks, Assembled
Pulling the Google reviews statistics into the seven numbers worth tracking: 20 reviews as the eligibility floor, 47 as competitive parity with the local top three, 4.0 as the rating cliff, 4.5 as the shelf worth climbing to, 90 days as the recency floor, 1 day as the response-time target, and 100% as the response-coverage target. Everything else in the surveys is context for those seven.
The reason this set is short is that the Google reviews statistics published each year mostly restate the same consumer behaviour from different angles, and the restatements do not change what you do on Monday. The floors do. Pick the one you are failing, fix that one, then recheck.
One honest caveat before you build a plan on any of this. BrightLocal’s survey is 1,002 US consumers on a representative panel, which is solid for direction and soft for precision; the difference between 47% and 44% in a sample that size is noise, and a figure quoted to the decimal point in a roundup post has usually been laundered through three other roundups first. Treat the thresholds as real and the second decimal as decoration. Then go and look at your newest review date, which is the one number nobody needs a survey to check.