Technology reviews are the main information source for most purchasing decisions, and they're produced within a set of constraints that affects what appears in them.
None of this makes reviews useless. It does change how to read them.
The access relationship
Most reviews are based on units supplied by manufacturers before release.
That access is valuable and revocable. A publication that reviews harshly may find future units arriving late or not at all, which affects its ability to publish at the moment when readers are looking.
This doesn't mean reviews are dishonest. It creates a structural pressure, and its effect is generally visible in tone rather than in content — criticism framed gently, negatives placed after positives, and reluctance to reach a clear negative conclusion.
The practical implication: attend to what a review says rather than to its overall tone. A negative buried in a paragraph of qualification is still a negative.
Embargo effects
Reviews are typically published simultaneously at an embargo time, meaning everyone had the same limited period with the product.
Which means launch reviews are based on days or weeks of use, not months. Anything that emerges over longer periods — battery degradation, durability, software problems, whether a company supports the product — cannot be in them.
Long-term follow-up reviews are considerably more informative and are published when nobody is searching, so they get read far less.
Searching for coverage several months after release, rather than at launch, is one of the more useful adjustments available.
Comparison problems
Reviews compare against what the reviewer has to hand, which is generally current flagship products.
Which means a product is assessed against the best available rather than against what the reader currently owns. A phone described as an incremental improvement over last year's model may be a substantial improvement over a four-year-old device.
Reviewers also use devices differently from most people — testing extensively, switching frequently, focusing on new features. Their experience of what matters is not representative.
Benchmarks and what they measure
Synthetic benchmarks provide comparable numbers and correspond imperfectly to actual use.
Sustained performance frequently differs from peak. A device that scores well in a short benchmark may throttle under sustained load, which matters for real work and not for the benchmark.
Manufacturers have been found detecting benchmark applications and altering behaviour, which has produced scandals periodically.
And the relationship between benchmark scores and perceptible differences is weak above a certain level. A device scoring meaningfully higher may feel identical in ordinary use.
What reviews systematically underweight
Software support duration. Determines how long a device remains usable and safe, and is rarely a prominent factor in scoring.
Repairability. Affects total cost of ownership and lifespan.
The company's track record. Whether previous products were supported, whether services were discontinued.
Long-term reliability. Cannot be assessed at launch and is frequently the most important factor.
Ecosystem lock-in. The cost of leaving, which accumulates invisibly.
Affiliate arrangements
Most technology publications earn commission on purchases through their links.
This is generally disclosed and creates an incentive towards recommending purchase rather than towards recommending keeping what you have.
Which partly explains why reviews of incremental updates rarely conclude that the improvement doesn't justify replacing a working device — a conclusion that would frequently be accurate.
Sponsored content
A distinct category requiring disclosure in most jurisdictions, with compliance varying.
The signals are recognisable: uniform enthusiasm across multiple outlets appearing simultaneously, identical framing, absence of comparison, and a focus on features rather than on whether to buy.
Video content has particular issues, since disclosure may appear briefly at the start of a long piece.
How to read more effectively
Read several, from different types of source. Enthusiast publications, general technology sites, and specialist communities have different biases.
Search for the product name plus a problem term. Owner forums surface issues that reviews cannot, because they aggregate experience over time and across many units.
Look for long-term follow-ups. Frequently the most useful content available.
Weight criticism heavily. Given the structural pressure towards positive framing, a clearly stated negative is a strong signal.
Check what wasn't tested. Reviews are explicit about scope more often than people notice.
Consider whether you need it. The question reviews cannot answer and rarely raise. A review tells you whether a product is good, not whether replacing a working device is a sensible use of money.
Community sources
Worth naming specifically because they answer questions reviews structurally cannot.
Product-specific forums and discussion communities aggregate experience across many units over long periods, which surfaces reliability problems, firmware issues and manufacturing variation that no single reviewer would encounter.
The bias runs the other way — people post about problems more than about things working — so a forum reads more negatively than reality. Adjusting for that, the pattern of which problems recur is genuinely informative.
Searching for the model number alongside a specific concern is frequently more useful than reading another review.