On-Chain Reputation vs Review Fraud: A Freelancer Benchmark

TL;DR

• Off-chain fake-review estimates run from single digits to over 20%, depending on unit and method.
• Paid fake app reviews averaged $1.73 to $4.24 in 2017.
• A 2025 study still found 3,551 reported farming addresses behind a $1,000 volume gate.
• In an illustrative model, a 10% fee makes a self-dealt $500 contract cost $50.

On-chain reputation is not fraud-proof, and no measured fraud rate for it has been published. Off-chain fakes are cheap and sold openly; an on-chain record is only as hard to fake as it is costly to write.

Each figure below keeps its own unit, year label and method, so this on-chain reputation fraud rate benchmark compares inputs, not one score.

What do measured off-chain review-fraud rates actually show?

They show a range. The FTC's 2024 rule record says estimates "vary by source" and "range from the single digits to over twenty percent," and that at least three studies found approximately ten percent of products or businesses have some manipulated consumer reviews. That 10% counts products or businesses, not individual reviews.

Four numbers, three units

• 16% of restaurant reviews filtered by one platform, a fraud proxy in a 2015 working paper.
• 11% to 15% of reviews estimated fake across 3 product categories on 9 e-commerce platforms used by UK consumers (products, not services), per a 2023 UK government-commissioned model.
• About 10% of products or businesses, the FTC summary.
• 121 of 1,550 sampled micro-task marketplace gigs (7.8%; the paper prints 6%) sold manipulation, 2013 data.

Why can't these studies be merged into one fraud rate?

Because units and methods differ: reviews, products or businesses, and gigs, measured by a filter, a classifier or a sample. The restaurant-review study found roughly 16% of reviews filtered and treats that as a fraud proxy, so 16% filtered is not 16% proven fake.

Why weak profiles tempt fraud

The same study found restaurants more likely to commit review fraud when their reputation is weak, with few reviews. A new freelancer profile sits in a similar spot.

Why is an off-chain review so cheap to fake?

Because a market sells them. In April 2017 data from 34 paid review providers (of 43 studied), fake reviews averaged $1.73 to $2.14 on one mobile app store and $3.41 to $4.24 on the other; some review-exchange portals paid reviewers, typically $0.20 to $1.50 per review. These are not 2026 prices.

A micro-task freelance marketplace sold them

In July and August 2013 data, 121 of 1,550 sampled gigs on a micro-task freelance marketplace were crowdturfing tasks sold for $5.

Social-media groups trade them

From March to October 2020, about 23 social-media groups a day trading fake e-commerce reviews each averaged 568 requests per day. The payoff is a significant but short-term rating boost.

How does an on-chain record change the cost to fake, and where does it fail?

A chain makes a record public, not honest. Cost is the lever that can deter fakes. In a 2025 study, one airdrop required at least two bridge transactions and $1,000 or more in volume, yet drew 3,551 reported attacker addresses; a second airdrop, whose gate the study does not state, drew 7,681. These are counts, not rates.

What a public record does and does not fix

Collusion survives auditability: when one person controls both sides of a contract, the record looks clean (see the attack vectors behind cheap wallets and self-dealing). In one e-commerce study, product-reviewer network features were more costly for sellers to manipulate than review text or timing: a hint, not proof, for on-chain signals.

How much would one fake approved contract cost (illustrative model, not a FiduWork mechanism)?

In this model, a self-dealt $500 approved contract under a 10% fee sends $50 to fees, roughly 12x to 29x the 2017 average price of a paid fake app review. That is original arithmetic on mixed-year inputs, not a platform measurement or a fraud rate.

FiduWork has not disclosed what event writes a reputation record, so the record-writing inputs below are assumptions, not platform rules.

Inputs, arithmetic and exclusions

• Assumed: records are written only on approved, fee-bearing contracts, and one attacker controls both wallets.
• Disclosed: a 10% flat fee on approved work. Assumed: a $500 fake milestone.
• Arithmetic: $500 x 10% = $50 per contract. $50 / $4.24 = 11.8 and $50 / $1.73 = 28.9. Against a $5 gig from 2013 data, it is 10x.

At a $50 fake milestone the fee is $5, the same as the 2013 gig. A fee gate bites only if small contracts carry little weight, and rating weights are not disclosed. The model excludes gas (a separate, per-transaction cost), locked-capital time value, identity or credential checks (not disclosed) and inflation since 2013 and 2017.

How do off-chain and on-chain trust signals compare side by side?

On unit price the gap is wide but conditional: a crowdturfing gig cost $5 in 2013 data, versus $50 in fees per $500 fake contract in the model. On prevalence there is nothing to compare: no fraud rate for on-chain freelance reputation, FiduWork's included, has been published.

• Measured manipulation figure · Consumer review and app-store platforms (off-chain, mixed sources): About 10% of products or businesses; estimates from single digits to over 20% (FTC 2024 rule record) · Micro-task freelance marketplace (off-chain): 121 of 1,550 sampled gigs (7.8%; paper prints 6%) sold manipulation, 2013 data (arXiv 1406.0574) · Open on-chain airdrop records (2025 study): 3,551 reported attacker addresses in a volume-gated airdrop; 7,681 in a second airdrop whose gate the study does not state; counts, not rates (arXiv 2503.14316) · Approval-gated on-chain record (illustrative model): No published rate; not measured
• Unit price to fake one signal · Consumer review and app-store platforms (off-chain, mixed sources): $1.73 to $4.24 average per paid app review, April 2017 data (arXiv 1904.12607) · Micro-task freelance marketplace (off-chain): $5 per gig, 2013 data (arXiv 1406.0574) · Open on-chain airdrop records (2025 study): 2+ bridge transactions and $1,000+ volume in the first airdrop, plus gas; second gate not stated (arXiv 2503.14316) · Approval-gated on-chain record (illustrative model): $50 per $500 contract under the 10% fee (model; FiduWork pricing page)
• How fakes are caught · Consumer review and app-store platforms (off-chain, mixed sources): Platform filter, used by researchers as a fraud proxy (SSRN 2293164) · Micro-task freelance marketplace (off-chain): Classifier at 97.35% accuracy in a study (arXiv 1406.0574) · Open on-chain airdrop records (2025 study): Reported-address lists published on GitHub (arXiv 2503.14316) · Approval-gated on-chain record (illustrative model): FiduWork rating dispute route: not disclosed
• How long a fake boost lasts · Consumer review and app-store platforms (off-chain, mixed sources): Significant but short-term (SSRN 3664992) · Micro-task freelance marketplace (off-chain): Not measured · Open on-chain airdrop records (2025 study): Not measured · Approval-gated on-chain record (illustrative model): Not measured
• Who can write the signal · Consumer review and app-store platforms (off-chain, mixed sources): Reviewers, including paid ones (arXiv 1904.12607) · Micro-task freelance marketplace (off-chain): The gig seller, for any buyer (arXiv 1406.0574) · Open on-chain airdrop records (2025 study): Any wallet that meets the gate (arXiv 2503.14316) · Approval-gated on-chain record (illustrative model): FiduWork: not disclosed

Reading the table without over-reading it

Columns don't share units. On-chain figures are counts, and "not measured" means no Tier-1 figure exists, not zero.

What should a freelancer do with this benchmark?

Never buy or swap reviews. US rules in force since 21 October 2024 ban buying fake reviews, treat an individual who sells services as a "business" and a review given "in exchange for" another review as a purchase, and star-only ratings count as consumer reviews too.

A four-point check for any reputation score

1. What event writes a record?
2. What cost sits behind each record?
3. Do bare star ratings count?
4. Can disputed work change the record?

Answer each from the platform's docs, its contract on a block explorer and one small test contract; a blank means unknown, not yes. Ask these of any platform, FiduWork included, and read an earlier explainer on wallet-bound credentials.

Platform Signal. FiduWork's BETA runs on Sepolia with 1,200+ freelancer profiles and a 10% flat fee on approved contracts. What writes, signs and weights a reputation record is undisclosed, so the cost figure above is a model.

Compliance corner. In the United States, the FTC's 16 CFR Part 465, effective 21 October 2024, makes it an unfair or deceptive practice for a business to write, create, sell or purchase fake consumer reviews (16 CFR 465.2); its definitions treat an individual selling services as a business and a review given in exchange for another review as a purchase (16 CFR 465.1). In the EU, Directive (EU) 2019/2161 added points 23b and 23c to Annex I of Directive 2005/29/EC, blacklisting submitting or commissioning false consumer reviews, and claims that reviews come from consumers who actually used or purchased the product without reasonable and proportionate checks, applying from 28 May 2022 (EUR-Lex text). Other jurisdictions, including the UK, have their own rules. This is general information, not legal advice.

Where to go next on FiduWork

Next: how DID reputation maps to the W3C VC data model, how a 72-hour juror vote works and landing your first Web3 contract.

Apply as a Verified Freelancer. Take on milestone work on the Sepolia beta and build your record from your own client work, not bought reviews. Open the app.

Sources

1. FTC final rule on consumer reviews, 89 FR 68034 (2024)
2. 16 CFR 465.1, definitions
3. 16 CFR 465.2, fake consumer reviews
4. Directive (EU) 2019/2161, EUR-Lex
5. Restaurant-review filtering study (2015)
6. Micro-task crowdturfing study (2014, 2013 data)
7. Paid fake app-review study (2017 data)
8. Fake review buyer networks (PNAS 2022)
9. Market for fake reviews (2022)
10. Airdrop Sybil measurement (2025)
11. UK fake online reviews research (2023), gov.uk

Update Log

• 2026-10-21: Published.