Trust and Safety in Marketplaces

By Brian Tr · Updated Oct 7, 2026

Strangers trade on a marketplace. Neither side can inspect the other before money moves, and a small share of the people who show up will try to cheat, steal, or hurt someone. Trust and safety (T&S) is the function that makes the honest majority willing to transact anyway.

It's often treated as a cost center bolted on after growth. In a marketplace, trust is the product. If buyers don't believe the listing is real, the network effects stop compounding.

This article defines the function, lays out the "trust stack" most marketplaces assemble, maps the main abuse types, separates prevention from detection and response, summarizes the legal duties that now exist, and suggests working metrics. We won't go deep on reviews or buyer protection, which have their own articles.

What trust and safety means

The Trust & Safety Professional Association (TSPA) describes it by its work. On its what-we-do page, the association says technology companies needed to determine which content and behaviors are appropriate and which aren't, and that the teams that handle this responsibility often fall under the general term "trust and safety." Its About page describes members as professionals who develop and enforce principles and policies that define acceptable behavior and content online.

For a marketplace, that broad definition narrows into something concrete. T&S is the set of policies, people, and systems that:

• decide who is allowed to join and what they're allowed to sell,
• stop or catch bad behavior before it hurts someone,
• resolve what goes wrong when something slips through, and
• keep the evidence and records regulators now expect.

It overlaps with fraud, support, and legal teams. T&S is the umbrella that coordinates them so a bad actor can't slip between teams.

Why marketplaces need it: the lemons problem

The economics are old. In his 1970 paper The Market for "Lemons", George Akerlof argued that where sellers know quality and buyers don't, dishonest dealings tend to drive honest dealings out of the market. He adds that the cost of dishonesty isn't only the amount by which a purchaser is cheated: it also includes the loss from driving legitimate business out of existence.

That's the case for T&S in one sentence. Every fake listing hurts the buyer who falls for it, and it also hurts the honest seller whose listing now looks suspicious. Left alone, the good supply leaves first.

A later strand of research looked at how online reputation could help. Resnick and Zeckhauser's study of eBay's reputation system describes reputations transmitted from person to person as a way to deter moral hazard and discourage entry by bad types in markets where players repeat transactions but rarely with the same player. Reputation is one layer of the stack below, and it's a strong one. It isn't enough alone, because it only works after someone has been burned.

The trust stack

Most mature marketplaces layer six controls. No single layer is sufficient, and each covers a different failure.

Identity verification

Verification answers a simple question: is this a real, identifiable person or business? How hard you ask depends on risk. Airbnb's identity verification help page says every host, new co-host, and booking guest must be identity verified, and that it may request a photo of a government ID plus a selfie for comparison. It also says hosts can't publish a listing until verification is complete. That's a strong stance, and it fits a category where people stay in each other's homes.

Few marketplaces need that level for every user. A low-value goods marketplace might verify a phone number at signup and escalate to ID checks as volume grows. We cover the seller side in seller onboarding.

Reputation

Reviews and ratings are the most visible trust signal, and also the most gameable. They're covered in their own article, so we'll only note the limit here: reputation reports on the past, so it can't stop a first-time scam.

Payments and escrow

Holding funds until the buyer confirms receipt, or until a service is delivered, removes the incentive for many scams and keeps the transaction visible to the platform. The mechanics are in marketplace payments.

Policy and enforcement

A policy nobody enforces is decoration. Good policies are specific, published in plain words, and tied to a ladder of responses: warning, listing removal, temporary suspension, permanent ban. Instant bans punish honest mistakes, while endless warnings reward persistent abusers.

Dispute resolution

When a transaction goes wrong, buyers and sellers need a route that isn't "email support and hope." Typical designs let the two parties try to resolve it directly, then escalate to the platform. Write the decision rules down, or sellers will assume they're arbitrary.

Insurance and guarantees

Some losses will always get through. Guarantees and insurance turn a rare disaster into a bounded cost, but a promise you can't afford to keep is worse than none. Guarantees have their own article.

The main abuse types

Different marketplaces face different mixes. These are the recurring categories.

Fraud against buyers. Fake listings, non-delivery, bait-and-switch, and account takeovers where a hijacked seller account lists goods it will never ship. Payment fraud, such as stolen cards, also hits sellers through chargebacks.

Fraud against sellers. Buyers who claim non-receipt falsely, return a different item, or pay with stolen credentials.

Counterfeit and prohibited goods. Fake branded goods, stolen property, unsafe products, and items banned by law or by your policy. This is where regulators are most active, and we'll cover the rules below.

Fake accounts and manipulation. Bots and sockpuppets creating fake reviews, inflating listings, or evading earlier bans.

Off-platform leakage. Users moving the deal elsewhere to avoid fees or to defraud, which strips away the platform's protections.

Harassment and abuse. Abusive messages between buyers and sellers, discrimination, and threats.

Physical safety. For in-person services and meetups (rides, home stays, local services, goods handed over in person), the risk isn't only financial. It includes assault, theft, and unsafe conditions. Verification, in-app communication, and clear incident reporting carry more weight here than in a pure e-commerce marketplace.

Treat these as separate problems with separate numbers. A single "bad content" bucket hides that counterfeit goods and harassment need different signals and responses.

Prevention, detection, response

A useful way to organize the work is by when it acts.

Prevention is the cheapest per incident but the most expensive to apply to everyone. Every extra step at signup loses some honest users. The standard approach is to match friction to risk: light checks for low-value activity, stronger ones when risk signals rise. That balance is a growth decision as much as a safety one, and it ties directly to supply quality.

Detection combines automated rules, machine-learning models, and people. Automation scales, but it misses new tactics and makes mistakes, so human review of edge cases matters. User reports are a detection channel too, and they need to be easy to find.

Response is where trust is won or lost. Fast, consistent action tells everyone the rules are real, and fair appeals tell honest users a mistaken flag isn't the end.

What the law now requires

Trust and safety used to be mostly voluntary. In several regions it's now a legal duty, at least for marketplaces selling to consumers. We'll summarize two regimes. This is a reference, not legal advice, and the details change, so read the primary texts for your situation.

EU Digital Services Act

Regulation (EU) 2022/2065, the Digital Services Act, has a section specific to online platforms that let consumers conclude distance contracts with traders.

• Traceability of traders (Article 30). The platform must ensure that traders can only use it to offer products to consumers in the Union if, beforehand, it has obtained information including the trader's name, address, telephone number, and email address; a copy of an identification document or electronic identification; the trader's payment account details; the trade register and registration number where applicable; and a self-certification committing to offer only products that comply with Union law. The platform must make best efforts to assess whether this information is reliable and complete, and must suspend traders who fail to correct inaccurate information after a request. It must store the information securely for six months after the contract with the trader ends.
• Compliance by design (Article 31). The platform's interface must be designed so traders can provide the information needed to comply with their legal obligations, including product identification and, where applicable, safety and labelling information. After allowing a trader on, the platform must make reasonable efforts to randomly check an official, freely accessible, machine-readable database to see whether offered products have been identified as illegal.
• Reporting and complaints. Hosting services must provide easy-to-access, user-friendly notice and action mechanisms (Article 16). Online platforms must give users an effective internal complaint-handling system to contest moderation decisions, for at least six months after the decision (Article 20), and notices from designated trusted flaggers must be given priority (Article 22).

US INFORM Consumers Act

The FTC's guidance on the INFORM Consumers Act, effective June 27, 2023, applies to online marketplaces where "high-volume third party sellers" offer new or unused consumer products. According to that guidance:

• A high-volume seller has, in any continuous 12-month period during the past 24 months, had 200 or more separate sales of new or unused consumer products and $5,000 or more in gross revenues on the platform.
• The marketplace must collect bank account, contact, and tax ID information, verify it, and require sellers to certify it at least once a year.
• Sellers with $20,000 or more in annual gross revenue on the platform trigger disclosure duties in listings or order confirmations.
• Non-compliant sellers must be suspended, and listings must carry a clear way for consumers to report suspicious conduct.
• The FTC page cites civil penalties of $53,088 per violation (the figure on the page as of this article's date).

The FTC states the purpose plainly: to add transparency and deter criminals from acquiring stolen, counterfeit, or unsafe items and selling them through marketplaces.

The common thread: verify who the seller is, keep records, give users a way to report, and act when information is wrong.

The scale of the problem

Fraud losses are large, and growing, at least as reported to US authorities. According to the FTC's March 2026 testimony announcement, consumers reported $15.9 billion in fraud losses in 2025 across 3 million fraud reports, up from 2.6 million reports and over $12 billion the year before. Imposter scams were the most frequently reported, and investment scams caused the largest losses.

These are consumer-fraud numbers across all channels, not marketplace-specific figures. They describe the environment your users live in, not your own loss rate.

Measuring trust and safety

T&S is hard to measure because success looks like nothing happening. The following are working definitions, not industry standards. Few have an agreed benchmark, so define each one carefully before comparing across teams or periods, and set targets from your own data.

Two cautions apply.

1. Track both errors. If you only measure how much bad content you removed, you'll over-enforce. Pair it with false positives and appeal outcomes. A system that never wrongly bans anyone usually misses abuse, and one that bans everyone suspicious drives out honest users.
2. Segment. Blend all abuse into one number and you'll miss that counterfeit listings are surging in one category or that a single country accounts for most chargebacks.

Pair these with seller churn after enforcement, since heavy enforcement can shrink supply. Read the trust stack alongside value-added services, since verification badges, protection plans, and similar features can be both safety features and products.

Where managed models fit

If your marketplace is the managed kind, with people vetting jobs and providers, a lot of this happens by default. See managed marketplaces. For why users try to move deals off the platform, see disintermediation.

Key Facts: Trust and Safety in Marketplaces

• The TSPA describes trust and safety as the general term for teams that decide which content and behaviors are appropriate on a platform and handle enforcement.
• Akerlof's 1970 paper argues that dishonest dealings tend to drive honest dealings out of a market, and that the cost includes the legitimate business driven out.
• Article 30 of the EU Digital Services Act requires online platforms to obtain identification, payment account, and self-certification information from traders before they can offer products to consumers in the Union, and to make best efforts to assess whether it is reliable.
• The FTC guidance says the INFORM Consumers Act covers sellers with 200 or more sales and $5,000 or more in gross revenue in a 12-month period, and requires marketplaces to collect, verify, and certify their information.
• The FTC reported $15.9 billion in consumer fraud losses for 2025, up from over $12 billion in 2024.
• Airbnb's help page says every host and booking guest must be identity verified, and hosts can't publish a listing until it's complete.

Related Reading

• Marketplace Supply Quality
• Marketplace Payments
• Marketplace Seller Onboarding
• Managed Marketplace
• Marketplace Value-Added Services
• Disintermediation: Why Users Go Around the Platform
• Reviews and Rating Systems in Marketplaces
• Guarantees and Buyer Protection in Marketplaces