Data

Online review trends 2026: what one year of data shows, like for like

24 October 2026 · JustReview

Measured on the same 549 businesses with a Google connection that stayed alive the whole time, customers wrote 0.9% more Google reviews between October 2025 and September 2026 than in the twelve months before. That is the most defensible answer we have to what changed in online reviews this year: the total stopped growing, after rising 18.5% the year before, and every quarter of 2026 came in slightly under 2025. Two louder numbers from the same database point in opposite directions, imports up 31% and reviews written down 18%, and neither of them says anything about customers. Under the flat line, 300 of those 549 owners saw fewer new reviews and 208 saw more.

Three counts that answer three different questions

We copy reviews from 54 source identifiers into one table for 3,825 accounts, and on 2026-10-24 that table held 1,407,542 rows. Each row carries two dates: the day the customer wrote the review and the day we stored it. Comparing October 2024 to September 2025 with October 2025 to September 2026 on both dates, plus one like-for-like panel, produces three headline numbers that disagree completely.

The first line is our own activity. When a business connects a source, we fetch its whole history, so a new customer with a ten-year-old profile adds ten years of reviews in one afternoon. In the latest year 42.6% of everything stored had been written before the window even began. Etsy is the clearest case: 12,553 rows stored, against 4,796 Etsy reviews actually written in the window. Booksy stored 8,940 and saw 2,183 written. Anyone who reports growth in “reviews processed” is mostly reporting new customers.

The second line looks like a market in retreat, and the third line says it is not. The gap between them is explained in the next two sections, and almost none of it involves a customer deciding not to write.

Where the 18% went: expired connections

Google accounts for 31,417 reviews written in the first year and 26,718 in the second, a drop of 4,699. Splitting every Google profile in the database by the state of its connection today accounts for all of it, and then some.

The 1,802 expired profiles lost more reviews than the whole source did. An authorisation can lapse for many reasons, from a changed password to a removed manager, and from that day our copy stops. The customers kept writing on Google; we simply stopped seeing it. Facebook shows the same mechanism at a larger ratio: 1,542 of 1,838 Facebook connections (83.9%) are marked expired, and Facebook reviews written in the window fell from 1,214 to 450. This is the same failure we described in which review sources still work in 2026, seen this time as a false trend line rather than a dead widget.

The 340 profiles connected during the period are the opposite trap. They grew 31.4%, most likely because people connect a review widget when reviews are flowing, not when they are drying up. A panel that keeps adding new customers will always look like a growing market.

eBay dates are import dates

eBay was our largest source by reviews dated inside both windows, and it cannot be compared year over year at all. Of the 30,941 eBay reviews dated between October 2025 and September 2026, 24,556 (79.4%) sit on just 36 seller-days that carry 100 or more reviews each. A year earlier the share was 67.4%. The largest single cluster is one seller with 3,963 reviews, every one of them dated 2026-07-27, which is the day we imported them.

That is not a buyer pattern. No seller collects four thousand pieces of feedback on one Monday. When the date a platform hands over is missing or coarse, a stored copy ends up with the fetch date, and from then on every chart built on it measures our import schedule. The weekly rhythm we found in eBay data in our online reviews statistics for 2026 has the same cause. Google returns its own timestamp for every review, so Google dates do not depend on when we fetch them, which is why the rest of this article leans on Google.

The same 549 Google profiles, quarter by quarter

The panel that answers the actual question is narrow on purpose: profiles connected before October 2024, with an authorisation that is still valid on 2026-10-24, and at least one review written in either year. Google’s owner API returns a profile’s full history, so a connection that broke for a few weeks and was restored still has every review, which makes this panel complete in a way the raw corpus is not.

Five quarters of growth, then three straight quarters a little below the year before. Over full years the panel wrote 13,478 reviews from October 2023 to September 2024, 15,966 in the following year and 16,113 in the latest one. Two cautions apply. A panel defined by who is still connected today leans towards survivors, which flatters the later years, so the real picture is, if anything, slightly weaker. And some of these businesses were young in 2023, which inflates the early growth. Neither caution changes the shape: the inflow stopped rising around the turn of 2026.

Flat total, shrinking middle

A flat total does not mean a typical business stood still. Of the 549 profiles, 300 collected fewer reviews in the latest year, 208 collected more, and 41 collected exactly the same number. Among the 500 that had at least one review in the first year, the median one ended up with 22% fewer, and 90 of them received none at all in the second year.

The gains were concentrated. The ten fastest-growing profiles added 1,809 reviews between them, which is 55.6% of everything gained across the panel. The single largest profile went from 1,424 to 1,837, and without it the panel shows -1.8% instead of +0.9%. So the honest summary is not “flat”, it is “a handful of busy businesses offset a slow decline everywhere else”, which matches what we measured from another angle in review velocity.

One table you will be tempted to build is growth by size, and it is a trap.

It looks like small businesses are booming and mid-sized ones are collapsing. Mostly they are not. Grouping by last year’s count puts every business that had a lucky year into a higher bucket and every unlucky one into a lower bucket, and next year both drift back towards their usual level. Statisticians call it regression to the mean, and it produces exactly this staircase on data where nothing changed. Any report that segments review growth by last year’s volume without correcting for it is showing you the method, not the market.

Ratings moved, slightly, upwards

On the same panel the share of five-star Google reviews went from 91.95% to 92.77%, the share of one- and two-star reviews fell from 3.56% to 2.96%, and the average went from 4.81 to 4.84. The change is small enough that a few businesses running a request campaign could produce it on their own. We would not call it a trend in consumer mood.

Other sources: too few like-for-like profiles to call a trend

We applied the closest equivalent to every other source: connected before October 2024 and still importing within the last 30 days. That condition leaves very few businesses per source, sorted here by how many.

These are anecdotes with a denominator. The Amazon line is one seller going from 126 to 1,067 reviews; the other eight mostly shrank. The only result consistent enough to mention is Google Shopping, where all four live connections declined. We would not publish any of the others as a platform trend, and we would be wary of anyone who publishes a platform trend without saying how many like-for-like businesses it rests on.

How to read your own year

The practical lesson is short. Count reviews written per month, per source and per location, and compare each month with the same month a year earlier. Before you treat a drop as real, confirm the connection behind it was alive for the whole period, because in our data an expired authorisation explained more of the decline than customers did. Ignore any number labelled “reviews processed” or “reviews collected” unless you know whether it counts the day of writing or the day of copying. And if a marketplace shows hundreds of reviews on one date, read that date as a delivery receipt.

Why we built JustReview around the review date

JustReview keeps both dates for every review, pulls full history from Google through the owner API, and shows the reviews on your website with each source labelled, so a reconnected Google profile gets back what was written while it was disconnected instead of showing a gap. The Google reviews widget is where most businesses start, and Booksy, Booking.com, Airbnb, Amazon, eBay, Etsy and other platforms can be added to the same widget. It displays the reviews you have; it does not create new ones, and nothing in this article suggests a widget changes how many customers write.

The free Basic plan covers up to 100 reviews from Google and Facebook. Marketplaces and booking platforms need a paid plan, and paid plans start at 9 EUR a month, as listed on pricing. You can create a free account and compare your own last twelve months with the quarters in the table above.