The product review, considered as a trust mechanism for retail commerce, was for approximately a decade one of the most reliable conversion levers available to a retailer. The review’s credibility derived from its independence — the customer wrote the review without obvious incentive, the retailer surfaced the review without obvious editorial interference, the prospective purchaser read the review with reasonable confidence that the substance reflected the customer’s actual experience. The mechanism is canonical, the implementations are widely available, and the conversion uplift the mechanism produced was, throughout the 2010s, sufficiently large that the review system’s presence on a retail catalogue was less of a competitive differentiator than its absence was a competitive disadvantage.
The mechanism has, in 2026, substantially collapsed. The collapse is not the result of a single decisive event but of an accumulation of pressures that have, individually, eroded different components of the trust contract; the cumulative effect is that the typical product review on a typical retail site is, on the qualitative survey data I have collected, treated by visitors as approximately as informative as a marketing claim from the retailer itself. The collapse is operationally consequential because the conversion uplift that the review system used to produce has, on the conversion data, declined alongside the trust; the system that was a reliable conversion lever in 2018 is, in many implementations in 2026, a piece of conversion theatre that produces no measurable improvement.
The pressures that have produced the collapse
The pressures are, in approximate order of magnitude, the following.
The first is the well-documented industry of paid review production, in which retailers acquire favourable reviews from organised networks of reviewers who post the reviews against catalogue items they have not, in any meaningful sense, purchased or used. The industry has existed in some form since the early days of online retail; its scale has, over the previous decade, increased to a point at which a non-trivial fraction of the reviews on the major retail platforms are produced by it. The platforms have, with varying degrees of success, deployed detection mechanisms; the mechanisms have, with varying degrees of success, identified and removed the most egregious cases. The detection-and-removal cycle has, however, never definitively cleared the platforms of the paid reviews; the cycle continues, and the visitor’s perception is that the cycle continues even when the specific reviews the visitor is reading are themselves authentic.
The second is the industry of incentivised review solicitation, in which the retailer offers an incentive (a discount, a loyalty-points credit, a follow-up product) in exchange for the customer writing a review. The practice is, in the strict sense, not a paid review; the customer’s experience and the customer’s review are both authentic. The practice nevertheless biases the review distribution towards positive reviews, on the well-documented grounds that customers receiving incentives are more likely to provide favourable reviews than those who are not. The bias is operationally consequential; the visitor, examining a review distribution that is mathematically inconsistent with the population’s likely satisfaction distribution, draws the inference that the reviews are systematically biased and discounts the substantive content accordingly.
The third is the well-documented practice of vendor moderation of reviews, in which the retailer selectively publishes favourable reviews and selectively suppresses unfavourable ones. The practice is, on the platforms that permit it, technically permitted under various editorial-discretion policies; the practice is also, from the visitor’s perspective, indistinguishable from outright fabrication of the favourable reviews. The visitor’s response is, in either case, to discount the entire review surface; the reviews that the retailer has not moderated are, by association, treated as untrustworthy as those the retailer has.
The fourth, and perhaps most consequential, pressure is the recent introduction of large-language-model review generation, in which the favourable reviews that previously required organised networks of human reviewers can now be produced at considerably lower cost via automated systems. The systems produce reviews that are, in the qualitative analysis of the trust-research literature, indistinguishable from authentic reviews on a per-review basis; the bulk indicators of fake-review production (improbably uniform sentiment, statistically detectable copy-paste patterns) are correspondingly more difficult to identify. The platforms’ detection capacity has, on the available evidence, not kept pace with the generation capacity; the proportion of fake reviews has, accordingly, increased rather than decreased over the previous eighteen months.
What visitors are doing instead
The visitor’s response to the trust collapse is, on the qualitative interview data, the substitution of alternative trust mechanisms for the product review. The substitutions vary by category and by demographic, but the dominant patterns are recognisable.
The first is the substitution towards user-generated video content, particularly on social platforms outside the retailer’s direct editorial control. A visitor who would, in 2018, have read the product reviews on the retailer’s site is, in 2026, more likely to search the product on TikTok or YouTube and watch a video review uploaded by a customer with no apparent commercial relationship to the retailer. The video content is, in the visitor’s perception, considerably more difficult to fabricate than the text review and is, accordingly, treated as a more reliable trust signal.
The second is the substitution towards specialist independent review sources — Wirecutter, RTINGS, the various category-specific review sites that have established editorial reputations. The visitor’s reliance on these sources has, on the qualitative data, increased substantially over the previous five years; the sources are treated as having editorial integrity that the retailer’s own review system is not.
The third is the substitution towards return-policy-based decision-making, in which the visitor proceeds with the purchase on the basis of the retailer’s return policy rather than on the basis of any prior trust signal. The pattern is operationally consequential because it shifts the trust burden from the pre-purchase review system to the post-purchase return system; retailers whose return policies are favourable convert visitors who would, in the previous trust environment, have been deterred by the absence of confidence-inspiring reviews. (For more on the return-policy dynamic, see Returns as a moat: why Asos got it right.)
What this implies for the retailer
The implication for retailers is, in some sense, an uncomfortable one. The product review system that has been a routine fixture of retail catalogues for over a decade is, in 2026, no longer a reliable conversion lever; the system continues to occupy substantial real estate on product pages and consumes ongoing operational attention; the system’s commercial contribution has, on the data, substantially eroded. The retailer’s available responses are limited and none of them are entirely satisfactory.
The first available response is the strict editorial integrity of the review system, with explicit verification of purchase, prohibition of incentivised reviews, and visible non-moderation of unfavourable reviews. The integrity is operationally available but is rarely commercially attractive; the visible presence of unfavourable reviews is, in the short term, a conversion regression that retailers are reluctant to accept even when the longer-term trust benefits are well-attested.
The second available response is the integration of the alternative trust signals into the retailer’s own catalogue — the surfacing of specialist review sources, the embedding of independent video content, the visible communication of the return policy at the moment of decision. The integration is operationally available but requires the retailer to acknowledge, in some implicit form, that the retailer’s own review system is no longer the dominant trust signal; the acknowledgement is, in some procurement conversations, an uncomfortable one.
The third available response is the strategic redeployment of the resources currently devoted to the review system to the trust signals that the visitor population is actually responding to in 2026. The redeployment is the recommendation that has emerged most consistently from the projects I have shipped; the resource availability is, on most retail budgets, the principal constraint, and the redeployment is therefore a question of opportunity cost rather than of additional spend.
An advisory close
The product review system has, in 2026, transitioned from a reliable conversion lever to a piece of legacy infrastructure whose continued operation is, on the data, producing a smaller commercial outcome than the resources allocated to it would suggest. The retailer’s task is not to abandon the system entirely — the residual value remains positive, and the visible absence of any review system would itself trigger trust signals that work against the retailer — but to recognise that the system’s commercial contribution has materially declined and to allocate the resources accordingly.
It is recommended that retailers auditing the trust mechanisms on their catalogues distinguish between the mechanisms that the visitor population is actively engaging with and the mechanisms that the visitor population has, in 2026, learned to discount; the distinction is decisive for which mechanisms warrant ongoing investment and which warrant reallocation of the resources to the alternatives the visitor is using instead.