What makes a review worth reading
A reader opening a review has a decision to make and wants someone who has already handled the product to help make it. A roundup reader wants a shortlist and a reason to trust it. In both cases the Search IntentThe goal behind a search query: to learn, to find a site, to compare or to act. is practical: which one should I buy, and will it suit me? Pages that answer with a rewritten specification sheet do not help, because the reader has already seen the specification.
The buying guides page records what Google's guidance on helpful content asks of this kind of page. It asks whether content provides original information and whether it shows first-hand expertise, for example from having actually used a product, and adds that readers trust product reviews more when they can see how many products were tested, what the results were and how the tests were run, with evidence such as photographs. That is a description of a method, not a style, and the rest of this guide is about having one.
Google's spam policies describe the opposite. The guide to Google's spam policies explains thin affiliation: affiliate pages whose product descriptions and reviews are copied from the merchant without original content or added value. Being an affiliate is not the problem. Original testing, comparisons and information about price and suitability are the value the policy contrasts with thin pagesPages that offer little original value of their own, such as near-empty templates, near-duplicates or copies with only a word or place changed..
This guide is for reviews of products the publisher does not make. Three neighbouring pages cover the other cases.
| Page type | Whose products | The trust problem | Where it is covered |
|---|---|---|---|
| Review or roundup by an independent publisher | Other makers' products | Showing the testing was real and the money did not pick the winner | This guide |
| Buying guide by a store | The store's own range | Helping shoppers choose without steering them to the dearest option | Buying guides |
| Comparison or alternatives page | The business's own product against rivals | Earning trust a vendor does not start with | Comparison and alternatives pages |
| Customer reviews on product pages | The business's own products, reviewed by buyers | Genuine reviews, no conditions on sentiment | Reviews and social proof |
Designing a test you can describe
Write the method before the first product arrives. A method written afterwards tends to describe whatever happened to favour the product the tester already liked.
Start from the reader. List the situations your readers buy for, in their words, and the questions they ask before buying: does it keep me dry on a long walk, does it fit over a jumper, can I pack it small. Each question becomes something the test checks. A test built from the specification sheet answers the maker's questions instead of the reader's.
Then decide, for each question, how it will be checked and recorded:
- Conditions. Where and how each product is used, kept as similar as possible across products so that the differences come from the products, not the weather or the day.
- Measurements. What can be measured, with what, and how often. Measure what readers care about rather than what is easy to measure.
- Judgements. What cannot be measured, such as comfort or ease of use, recorded by the same person in the same way for each product, with notes on why.
- Duration. How long each product is used before it is scored, the same for every product.
- Who tests. The named tester, their experience of the category, and anyone else whose judgement is recorded.
Publish the method, in plain words, on every review that uses it, or on a methodology page each review links to. The reader should be able to tell what was done without guessing. Where a product could not be tested in the same way as the others, say so in its entry rather than scoring it as if it had been.
Evidence: photographs, measurements and notes
Evidence is what lets a reader believe the test happened. The buying guides page names photographs as an example, and the E-E-A-T guide lists original images, data gathered and tests run, with dates, as evidence of experience. It also observes that for most product reviews, experience counts for more than formal expertise: the person who has used the product is the expert.
Keep three kinds of record from the first day:
- Photographs of each product in use, in the test conditions, taken by the tester, including the details readers ask about and the faults you find. Stock images and the maker's photographs show nothing about your test.
- Measurements in a sheet, one row per product per test, with the date and conditions. The published review quotes from the sheet, so every figure on the page can be traced.
- Notes written during or straight after each use, before the tester knows the scores. The quotable observations in a good review usually come from these notes.
On the page, show the evidence where the claim is made: the photograph of the seam that leaked beside the sentence that says it leaked. A gallery at the end of the review is better than nothing, but readers who scan, as the writing for the web guide explains most do, see the claims and skip the gallery.
Keep the raw records after publication. They answer readers' questions, settle a maker's complaint and make a re-test comparable with the first.
Scoring and choosing winners
A score makes a judgement comparable across products. Use a weighted scoring matrix and follow its discipline, which exists for exactly this problem.
- Weights before testing. Agree the criteria and their weights from the readers' questions before any product is used. The framework warns that once a team has a favourite, weights drift towards the qualities the favourite happens to have. Record the weights with a date.
- Must-haves as gates. A product that fails a requirement readers cannot do without, such as a jacket that lets water in, fails, whatever it scores elsewhere. A heavy weight is not a substitute for a gate.
- Anchors for each score. Write what a low, a middle and a high score look like for each criterion, so that the score measures the product and not the scorer's mood.
- Sensitivity. Re-run the totals with another plausible set of weights. If the winner changes, the decision turns on one judgement, and the review should say so.
- Close totals are ties. The framework's advice applies: a narrow gap is inside the margin of the judgements, so call it a tie and say which product suits which reader.
Choosing winners follows from the scores, and labels such as best overall, best for a tight budget or best for long walks each name the reader they serve. A winner chosen for any other reason, such as the commission it pays, the maker's advertising or a relationship with the brand, is not a review result, and it breaks the promise the method page makes. If commercial considerations ever influence what is included or how it is ranked, the reader needs to be told; most publishers decide instead that they never will, and write that into their editorial standards.
Structuring a roundup
A roundup is several reviews and a decision aid on one page, read mostly by scanners. Structure it for the reader who wants the answer first and the reader who wants to check it, in roughly this order.
- The short answer. The winners, each with the reader it suits and the one reason, at the top.
- The comparison table. Every product tested, with its scores on the main criteria, how it arrived and a link to its entry. A table readers can scan does more than paragraphs of praise.
- How we tested. A short version of the method, linking to the full methodology, the tester's name and the dates of testing.
- One entry per product. What it is, who it suits, what the test found, the evidence, the faults and who should not buy it.
- What we left out, and why. Products considered but not tested, or tested and dropped, with the reason.
- Disclosure. How products arrived and how the site earns, near the top and again beside the links.
- Change history. When the roundup was last re-tested and what changed.
Every product in the roundup was tested by the same method. Padding a roundup with products the team has not used, to cover more searches, turns a review into a list, and the untested entries are the ones the helpful content questions and the thin affiliation policy describe. A shorter roundup of products you actually tested is the honest version.
Write entries in the language of the writing for the web guide: descriptive headings that name the product and its verdict, the main point first, and faults stated as plainly as strengths. An entry with no faults reads as an advert.
Disclosure: how products arrived and how the site earns
Readers weigh a review differently when they know the product was free or the link pays the site. Two separate things need handling: what readers are told, and how the links are marked.
What readers are told. The E-E-A-T guide lists, under honesty, that reviews disclose how products were obtained, and under transparency, how the site makes money and what affiliate or sponsorship relationships exist on the page. The FTC's endorsement guidance, as the influencer contracts and disclosure guide describes it, expects material connections to be disclosed clearly and conspicuously, and covers gifts and affiliate links among common situations. The media and publishing guide records that affiliate links in reviews are advertising and that the UK advertising codes administered by the ASA require marketing to be obviously identifiable as such. In practice that means a plain statement near the top of the page saying how each product reached the tester and that the site earns when readers buy through some links, repeated where the links are, in words an ordinary reader understands. A link to a disclosure page in the footer alone is not where readers meet it. The exact wording is a question for the regulator's guidance in your markets and your adviser; this guide is not legal advice.
How the links are marked. The outbound links guide sets out Google's documentation on qualifying outbound links: sponsored for adverts and paid placements, with nofollowA link attribute asking search engines not to treat the link as an editorial endorsement. still accepted for paid links. It lists affiliate links, and links in reviews of products sent free where the arrangement included a link or the expectation of one, among the links that need the label, and notes that sending a product in return for a write-up with a link appears among the link spam policy's examples. Apply the attribute in the template or the affiliate link tool so that it does not depend on writers remembering.
| How the product arrived | What readers are told | How links to it are marked |
|---|---|---|
| Bought by the site | That the site bought it | Editorial links need no attribute; affiliate links carry sponsored |
| Lent by the maker and returned | That it was lent, and returned after testing | Affiliate links carry sponsored; as general practice, treat a loan that came with the expectation of a link like a gift |
| Gifted by the maker | That it was given free | Sponsored where the arrangement included a link or the expectation of one, and on affiliate links |
| Paid placement in the roundup | That the placement is paid, as advertising | Sponsored, and the page labelled accordingly |
Affiliate programmes have their own disclosure terms, and the affiliate compliance guide covers what advertisers expect of the publishers in their programmes. The tutorial on checking affiliate content for disclosure and claims shows the advertiser's side: how a brand samples published reviews and asks for fixes.
Two more rules protect the review itself. Review markup, the Schema MarkupStructured data vocabulary. that can show ratings in search, must reflect the real review visible on the page: the structured data guide records from Google's structured data policies that markup must describe visible content and not misrepresent it, and that self-serving reviews of your own organisation are not eligible. And every review is written by a person who did the testing. The AI claims and consumer protection guide puts it plainly: a generated review is a fake reviewA review not based on a genuine experience, such as one written by the business, paid for without the experience or bought from a service. whatever prompt produced it. AI tools can help organise notes or check the sheet; the experience has to be the tester's.
Keeping reviews current
Products change after the review is published: makers redesign them, change materials, raise prices or withdraw them. A review that describes a product the reader can no longer buy misleads the reader, however good the test was.
Give every review and roundup an owner and a review date in the content inventory, and watch for the changes that trigger an earlier look: a new model or version, a reader reporting a fault the review did not find, a product going out of stock for good, or a change in how the site earns from it. When a product changes in a way the test would notice, re-test it with the same method and the same weights, and compare against the original records.
Show the change on the page. A short dated note saying what was re-tested and what changed tells returning readers whether to read again, and the updated date beside the published date means the substance changed, not that someone re-saved the page. The guide to updating and correcting published content sets the house rules for update notes and corrections. A product that is withdrawn comes out of the roundup's winners with a note, rather than staying as a recommendation that leads to a dead link. When a business films its own reviews for YouTube, the guide to making product review and comparison videos adapts the test, the scoring and the disclosure to video, where corrections go in the description and a pinned comment.
A hiking gear site tests rain jackets (illustrative)
Puddlefern Outdoor Reviews tests rain jackets for a roundup
Puddlefern Outdoor Reviews, an invented hiking-gear review site, plans a roundup of rain jackets for walkers. Before the first walk, its tester writes the method and a weighted scoring sheet from readers' questions: staying dry 35 points, breathability on climbs 20, fit over layers 15, packed size 10, hood 10 and pockets 10, with any jacket that lets water through at the seams failing outright. It shortlists 9 jackets: 4 bought by the site, 5 lent by makers and returned afterwards. Each jacket is worn on the same 3 routes in rain over 6 weeks, with the same base layers, and the tester photographs each one wet at the shoulders, cuffs and hood and records notes after every walk. One jacket fails the seam gate in week 2 and is listed with the reason and a photograph. The top 2 finish 3 points apart, so the roundup calls them a tie and names the reader each suits. A statement at the top says which jackets were lent and that the site earns commission on some links; the same line sits beside the links, and every affiliate link carries sponsored through the site's link tool. Five months later the maker of the budget winner redesigns it; Puddlefern re-tests the new version on the same routes, finds the new hood smaller, lowers its hood score from 8 to 5, keeps it as the budget choice and adds a dated note saying what was re-tested and why.
Common mistakes
- Writing the method after testing. A method written afterwards describes the result rather than producing it.
- Reviews of products nobody used. Summarising the maker's description adds nothing, and with affiliate links it is the thin affiliation pattern.
- Stock photographs as evidence. The maker's images show the product, not your test. Use the tester's own.
- Weights changed after scoring. Moving the weights to fit a favourite makes the scores decoration.
- Winners chosen by commission. If the money picks the winner, the review is an advert with a review's layout.
- Padding a roundup. Untested products added for search coverage undermine the tested ones.
- Disclosure only in the footer. Readers meet the review and the links, not the footer; disclose where they are.
- Unqualified affiliate links. Every affiliate and gifted-product link carries sponsored, applied by the template.
- Ratings in markup that differ from the page. Review markup describes the real, visible review and nothing else.
- Reviews that never change. A redesigned or withdrawn product needs a re-test or a note, not silence.
