Back to feed

How to Read a Software Review Without Being Sold To

Six signals the writer actually used the product, six that it is a rewritten press release, and the five questions a review should answer but usually does not.

S
Samina
Aug 6, 2026 · 4 min read
Top Tech Gadgets and Software Reviews Worth Reading in 2026

Search for any software comparison and the first page is mostly affiliate content. That is not automatically bad — someone has to pay for the testing — but it changes what the article is optimised for, and it is worth being able to tell.

Here is how to read a software review in about ninety seconds and know whether the writer used the product, and whether the ranking reflects quality or commission.

Six signals the writer actually used it

  • A specific complaint. Not “the interface can feel cluttered” but “the export dialog buries PDF under a submenu”. Marketing material never contains a specific complaint, so a real one is hard to fake.
  • Original screenshots. With their own data in them, at an odd window size, ideally showing a state you would only reach by using the product for a while.
  • Numbers with a method. “Sync took 40 seconds on a 2 GB vault” beats “sync is fast”. If the number has no method, it came from a press page.
  • A named version. Software changes. A review that does not say which version or month it tested is either old or hypothetical.
  • Someone told not to buy it. Genuine reviews contain a paragraph explaining who should choose something else. Affiliate reviews rarely do.
  • Something that went wrong. Real use produces friction — an import that failed, a setting that had to be found in a forum.
Laptop showing software comparison articles
Someone pays for testing. The question is whether it shaped the conclusion.

Six signals it is a rewritten press release

SignalWhat it usually means
Every option is “great for” someoneAll of them are monetised
Pros and cons are perfectly balanced 3 and 3The cons were invented to look fair
Feature lists match the vendor’s site orderCopied from the pricing page
No pricing, or “contact for pricing”Writer never reached a checkout
Only vendor-supplied imagesProduct was never opened
The winner is also the highest commissionCross-check two or three articles and the pattern shows
One signal proves nothing. Three or more and you are reading marketing.

Affiliate links are not the problem

Testing software properly takes days. Somebody pays for that, and it is either the reader, an advertiser, or an affiliate programme. The third option is the least intrusive of the three.

What matters is whether the commission shaped the conclusion. Two things tell you quickly:

  • Does the article ever recommend something with no affiliate programme? Free and open-source tools pay nothing. A list that never mentions one, in a category where good free options exist, is filtered.
  • Is the disclosure at the top or buried? Not a legal question so much as a signal about how the site sees its readers.
Person reading reviews on a laptop
Three or more warning signals and you are reading marketing.

The questions a review should answer

Most reviews answer “what does it do”. The useful ones answer these instead:

  1. What does it cost in year two? Introductory pricing is not pricing.
  2. Who should not buy this? The most useful paragraph in any review.
  3. What happens to my data if I leave? Export format, and whether the export is actually usable.
  4. What breaks at scale? Every tool has a size where it slows down. Good reviews name it.
  5. How long until it is set up properly? Not installed — set up. Often the difference between an afternoon and a fortnight.

If a review answers three of those five, it was written by someone who used the thing.

A faster way to decide

Reviews are best for narrowing a field, not for choosing. Once you are down to two or three candidates, better sources exist:

  • Search the product name plus “migrating away”. People who left explain the real limits better than anyone still using it.
  • Read the changelog. Six months of releases tells you whether the team is fixing things or shipping features nobody asked for.
  • Look at the support forum, not the subreddit. The forum shows what breaks; the subreddit shows what enthusiasts love.
  • Then use it for two weeks. Nothing substitutes for this, and almost everything worth using has a trial.

For what it is worth, that is the standard we try to hold ourselves to here — specific numbers, named limits, and a clear statement of who should pick something else. Our comparisons of Notion vs Obsidian and Todoist vs TickTick both open by naming who should not choose the tool that wins.

Desk with laptop and notes comparing tools
Reviews narrow a field. Two weeks of use decides it.

Common questions

Are user reviews on app stores more reliable?

Different bias, not less. They skew to people who are delighted or furious, and both write immediately rather than after six months of real use. Sort by most recent and read the three-star ones — they are the most informative.

How old is too old for a software review?

Beyond about eighteen months, treat pricing and feature claims as unreliable. The judgement about who a tool suits usually ages much better than the specifics.

What about AI-generated reviews?

They read fluently and contain no specific complaints, no original screenshots and no numbers with a method — which is precisely what the checklist above catches. The signals have not changed; there is just more of the bad kind now.

Should I trust a review that says everything is good?

No. Every tool involves a trade-off, and a review that cannot name the trade-off has not found it yet.


Keep reading

One response to “How to Read a Software Review Without Being Sold To”

  1. […] Ask what it would look like if the claim were false. If you cannot describe that, it is not a prediction — it is an advertisement. The same test works on software reviews, which we covered here. […]

Leave a Reply

Your email address will not be published. Required fields are marked *