- Google's reviews system evaluates standalone articles that recommend or analyze something. Most of its published best practices describe evidence rather than writing quality.
- Google stopped announcing individual updates to the system in late 2023, so sites that fail the bar now decline without a headline to explain why.
- Six things function as visible proof of experience: original photographs, measurements you took, failures, named testers, dates, and plain disclosure.
- Testing pays twice now. It earns the ranking, and it earns the citation inside AI answers that never produce a click.
Search for the best anything right now and you will get twenty pages that read the same way. Same products, same order, same adjectives. Nobody who wrote them has held the thing.
Google knows this. It has been working on the problem since 2021 through a set of ranking systems built specifically to sort content written from experience from content written from other people's content. The guidance has been public the whole time. Most publishers have read it and carried on summarizing spec sheets anyway, because testing products costs money and summarizing them does not.
That calculation has changed. Here is what the systems actually ask for, what proof looks like on a page, and why the gap between testing and summarizing matters more every year.
The Reviews System, in Plain Terms
Google runs a ranking system aimed at review content. It started in 2021 as the product reviews system. In April 2023 Google renamed it the reviews system and widened its scope to cover services and other subjects rather than physical products alone.
The system evaluates first-party standalone content that recommends something, gives an opinion, or offers analysis. It does not evaluate customer reviews sitting in a widget at the bottom of a product page. It looks at the article you wrote.
Google stopped announcing individual updates to it in late 2023, saying improvements now ship on an ongoing basis. That change matters more than it sounds. There is no longer an update date to point at when traffic drops. The bar moves continuously, and sites that fail it decline without a headline to explain why.
Alongside the system, Google publishes a list of best practices for writing reviews. Several of them describe evidence rather than writing quality.
What the guidance actually asks for
- Evaluate from the reader's perspective, rather than from the manufacturer's.
- Demonstrate that you know the thing well, through your own use of it.
- Supply evidence such as visuals, audio, or other links documenting your own experience, which supports your expertise and backs up the authenticity of the review.
- Share quantitative measurements of how something performs.
- Explain what sets this option apart from the alternatives, and cover comparable options you did not pick.
- Give first-hand evidence whenever you call something the best for a purpose.
Read that list again and notice how little of it concerns prose. Most of it concerns whether you did the work.
Where Experience Fits in E-E-A-T
In December 2022 Google added a second E to E-A-T. Expertise, Authoritativeness and Trustworthiness became Experience, Expertise, Authoritativeness and Trustworthiness.
The Quality Rater Guidelines describe Experience as the extent to which the content creator has the necessary first-hand or life experience for the topic. The raters do not score your site directly and their ratings do not feed a ranking factor. What they do is tell you, in Google's own words, what Google is trying to measure.
For a review, experience means you used it. For a travel guide, it means you went. Analysis after the 2023 reviews update found a consistent pattern among sites that lost visibility: their content read as information gathered from around the web, with no evidence the writer had been anywhere near the subject.
What Evidence Looks Like on a Page
Experience is a claim until the page carries proof. Six things function as proof, and all six are visible to a reader, which is the point.
| Evidence | What it looks like | What it replaces |
|---|---|---|
| Original photographs | Your hand, your counter, your lighting, the scuff on the corner | Manufacturer renders and stock imagery |
| Measurements you took | Response time in seconds, battery life in hours, range in feet | Numbers copied from the spec sheet |
| Failures | What broke, what never worked, what you would not recommend now | Uniform praise that reads as sponsored |
| Named people | Who tested it, what qualifies them, a profile a reader can check | Staff bylines with no archive |
| Dates | When you tested, when you last looked again | Undated pages with stale prices |
| Disclosure | How you got the product and how you make money, stated plainly | A policy buried in the footer |
A number from a spec sheet is the manufacturer's claim. A number from your stopwatch is yours, and nobody else on the results page has it. That difference is the whole asset.
A Worked Example
Good Medical is a useful illustration, because its methodology is the article it never had to write.
The site reviews medical alert systems, mobility aids, fall-prevention gear and other aging-in-place products for older adults. Its methodology page leads with a line that frames the whole operation: it says the team does not review, it tests.
Each of those claims maps onto something the guidance asks for. The site says it buys every product at retail rather than accepting PR samples, which removes the incentive problem before it starts. It says products go to a panel of testers aged 62 to 89 who use them in their actual homes for weeks rather than in an office for an afternoon, which is first-hand experience by the strictest reading. It publishes measured figures, response times in seconds and battery life in hours, instead of repeating the manufacturer's numbers. It re-tests and re-ranks every product quarterly, because plans and prices change. It names its editor in chief, states her background in consumer product journalism, and says it runs zero sponsored placements. It also discloses the affiliate relationship plainly rather than burying it.
None of that is clever. It is expensive, and that is the actual barrier. Any competitor can copy the page structure in an afternoon and cannot copy hundreds of products bought at retail and lived with for weeks. The methodology page is doing double duty: it tells readers why to trust the recommendation, and it gives search engines the specific evidence the guidance asks for, in one place, in writing.
Why This Matters More Now
Two things have raised the stakes.
The first is volume. Generating a plausible 2,000 word roundup now costs almost nothing, so the supply of untested review content has gone close to infinite. When supply is infinite, the only thing with value is what cannot be generated. A stopwatch reading from a product sitting on your desk cannot be generated.
The second is AI search. When somebody asks an assistant which medical alert system to buy for their mother, the assistant assembles an answer from sources it treats as reliable and names them. Original measurements give a model something specific to cite. A page that paraphrases five other pages gives it nothing that the other five did not already provide, so it gets passed over.
The result is that first-hand testing now pays twice. It earns the ranking, and it earns the citation inside answers that never produce a click at all.
Need an agency that builds this properly?
Browse vetted content and SEO agencies that work from original research rather than rewritten roundups.
Compare agenciesIf You Publish Reviews, Start Here
Pick one category and test it properly
Depth in one subject beats thin coverage across forty categories written from a desk.
Write the methodology page before the reviews
State what you buy, how you test, how long for, who tests, how often you revisit, and how you make money. Link it from every review.
Put one original measurement in every review
One number nobody else has is worth more than another thousand words.
Photograph what you tested
Even badly. A real photo of a real unit outperforms a perfect render.
Say what you did not like
Every product has a flaw. Naming it buys you the credibility to make the recommendation.
Name your testers and date your pages
Both are trivial to add and both are checkable. Re-test on a schedule and show the updated date, because stale reviews decay quietly.
What It Costs
Worth being honest about this. Buying products at retail, shipping them to testers, paying for weeks of their time and repeating the whole cycle quarterly costs real money, and it will always lose a short-term cost comparison against generating the same page for nothing.
The reason to do it anyway is that the cheap version has no floor. Anyone can produce it, so it competes with everything, and it loses value as the supply grows. Testing builds an asset that compounds instead, because every quarter of measurements makes the archive harder to replicate.
That is the whole trade. One route is cheap now and worth nothing later. The other is expensive now and owns a category.
Frequently Asked Questions
It is a ranking system aimed at review content. It launched in 2021 covering product reviews and was widened in April 2023 to cover services and other subjects. It evaluates standalone articles that recommend, give an opinion on, or analyze something. It does not evaluate customer review widgets on a product page.
No. Google said in late 2023 that improvements to the reviews system now ship on a regular, ongoing basis and that it would stop posting individual update notifications. Ranking changes from this system no longer come with an announcement attached.
Using the thing yourself and showing it. Original photographs, measurements you took, problems you ran into, and named people who did the testing. Summarizing specifications and other people's reviews does not qualify, however well written the summary is.
You need evidence proportionate to the claim. A buying guide that recommends a best overall should carry first-hand support for that specific call. Testing three products properly and saying so beats claiming to have tested thirty without proof.
No. Affiliate links are a normal business model and Google has never treated them as a quality problem on their own. What causes problems is content written to place links rather than to help somebody decide, and undisclosed commercial relationships. Disclose clearly and test honestly, and the revenue model is irrelevant to the ranking.
Assistants assemble answers from sources they treat as reliable and name them. Original measurements give a model something specific to cite that no other page carries. A page that paraphrases five other pages offers nothing those five did not already provide, so it gets passed over.
Sources & Further Reading
- Write high quality reviews · Google Search Central
- Google releases April 2023 reviews update · Search Engine Land
- Google releases November 2023 reviews update · Search Engine Land
- Google's newest reviews update elevates real life experience · Amsive