Fake-review broker networks in 2026: an investigation

The 2026 broker market at a glance

Two years after the FTC final rule and the UK DMCC Act, the fake-review economy has not disappeared — it has professionalised. Cottage-industry sellers on freelancer marketplaces have been displaced by a smaller number of well-organised operations advertising on encrypted channels and open Telegram groups.

The professionalisation matters because it changes the correct enforcement posture. Chasing individual fake reviews one by one — the 2019 approach — is now essentially useless. Detection has moved to network-level analysis: temporal correlation, embedding similarity and reviewer-graph clustering.

The pricing above is the median for a UK-targeted five-star review with 'aged-account' delivery. Photo-attached reviews trade at roughly 3× that, and 'guaranteed publication' upsells (where the broker refunds if the platform removes the review inside 30 days) add another 40–60%.

Active broker operations mapped

41

UK-facing platforms targeted

17

Median price per five-star review

£11.20

Share of brokers offering 'guaranteed removal'

38%

Median network burst size

78 reviews in 11 days

Actions taken by ScoreReview UK in the period

3,914

How the brokers actually operate

The internal architecture of a professional broker in 2026 is roughly the same across the 41 operations we mapped. It has three layers: an account pool, a language layer and a distribution layer.

The account pool is the most expensive layer to build and the hardest to replace, which is why brokers guard it fiercely. Losing a pool costs months of work, and every enforcement action against a pool is a durable win.

The language layer is now almost entirely LLM-generated. This is a double-edged sword for the brokers: it lets them scale text production dramatically, but it also creates recognisable statistical fingerprints that specialised classifiers can pick up. We publish our detection rates by broker cohort each quarter in the enforcement log.

“The single most useful shift in the last 18 months has been treating the reviewer graph as the primary detection surface. Language classifiers alone were losing the arms race. Graph clustering flips it, because you can't fake a network history retroactively.”

The tell-tale patterns consumers can spot

You do not need a graph classifier to catch most brokered content. The observable patterns on a business profile are surprisingly consistent.

1. 1A histogram bar that spikes 5–10× above the surrounding weeks and then falls back to zero — the classic 'campaign' shape.
2. 2A cluster of reviewers whose profiles show 1–3 reviews total, all left inside the same 14-day window.
3. 3Reviews with no proper nouns: no staff name, no product name, no dates, no problem resolved.
4. 4A 'reply rate' that plummets on the campaign window because the business is not organised to respond to a burst.
5. 5A sector-atypical rating distribution: a plumber with 100% five-star and zero replies is more suspicious than a plumber with a 4.6 average and thirty replies.
6. 6Cross-platform overlap: the same reviewer handles appearing on three other platforms in the same week is a strong graph signal.

Where enforcement is still failing consumers

The DMCC Act 2024 and the FTC final rule closed most of the loopholes, but three gaps remain visible in our data set. Naming them here is deliberate — they are the priorities for 2026 policy work.

First, cross-platform data sharing remains voluntary. The 2025 industry MOU is a good start, but only a subset of platforms have signed, and consumer protection agencies do not yet have a formal role. A broker banned from four platforms can still trade on a fifth for months before detection catches up.

Second, 'guaranteed removal' services are not directly criminalised. The DMCC Act penalises the commissioning and publication of fake reviews, but the intermediate market — services that pretend to be reputation-management firms while operating as pay-to-suppress channels — sits in an enforcement grey area.

Third, small businesses commissioning fake reviews rarely face enforcement action. The CMA's 2025 case work has focused on marketplaces and platforms, which is appropriate as a priority but leaves the demand side of the market undeterred. Until commissioning is prosecuted at the SME level, supply will find willing customers.

None of these gaps is unsolvable. All three are visible in the enforcement logs of the platforms already doing serious work, which means the evidence base for the next policy iteration is already accumulating.

Conclusions and what to watch in 2026

Fake reviews are not going away. They are becoming more expensive to run and more detectable when run — which is a form of progress, even if it is not the same as elimination.

The correct consumer posture in 2026 is not to trust or distrust platforms wholesale, but to read profiles structurally: histograms, tier distributions, reply patterns, cross-platform consistency. The tools are increasingly available; the habit is still catching up.

The correct platform posture is to publish enforcement metrics unprompted, to sign cross-platform data-sharing agreements, and to invest in graph-level detection rather than defending the aging per-review classifier stack. This report exists partly to raise that bar publicly.