Compliance

Cloaking — the legal grey, and what compliance teams actually do

A working map of the cloaking landscape in 2026 — the hard-illegal, the regulatory grey, and the operationally tolerated. With FTC enforcement context, recent settlements, and what compliance teams at the major networks actually catch versus miss.

By Eyal RosenthalMay 6, 202616 min readAI-assisted research

Cloaking is the affiliate term for showing different content to different visitors — typically, one page to the ad network's reviewer or the search engine's crawler, and a different page to the actual incoming user from the ad. The word "cloaking" carries enough criminal-flavor that operators avoid using it on their resumes, but the technical practice spans a wide spectrum. Some forms are blatant fraud against ad networks and consumers. Some are mainstream practices that every ecommerce site uses (showing different prices to different geos, for instance). The shape of the landscape is poorly understood, even by people who have built businesses inside it. This piece tries to map it accurately, with citations.

A precise definition

When I say "cloaking," I mean: programmatically serving different content to different requests on the same URL based on properties of the request (IP, user-agent, geo, device, referer, time of day, cookie state, JS-execution profile). That is the technical mechanism. The ethical and legal status varies entirely with what content is being substituted and what the substitution is intended to deceive.

A few examples that are all technically cloaking but morally and legally very different:

Geographic price differentiation. Showing $9 to US visitors and €9 to EU visitors. Standard ecommerce. Fully legal.
Device-based layout switching. Showing a mobile-optimized page to mobile and a desktop-optimized page to desktop. Standard web design. Fully legal.
Crawler-vs-user differentiation for SEO. Showing a different rendered version of the page to Googlebot than to users to optimize for crawl efficiency. Borderline; subject to Google's Webmaster Guidelines, which forbid certain forms of cloaking but tolerate others.
Compliance-team detection. Showing a "white page" (compliance-friendly content) when the request appears to come from an ad network's compliance reviewer, and the "money page" (the real offer) when the request appears to come from a real user. This is the affiliate-marketing usage. This is the practice that gets people in trouble.
FTC-actionable deception. Substituting affirmative misrepresentations — fake reviews, fake countdowns, fake celebrity endorsements — that would trigger Section 5 of the FTC Act if seen by the consumer.

The first three are mainstream. The fourth is the operational reality of much affiliate marketing. The fifth is fraud.

What the FTC has said and done

The FTC's general authority under Section 5 of the FTC Act prohibits "unfair or deceptive acts or practices in or affecting commerce." A cloaked landing page that presents different content to compliance reviewers than to consumers, where the consumer-facing content is itself deceptive, is squarely within Section 5.

The FTC has not historically prosecuted cloaking as cloaking. They prosecute the underlying deception. That distinction matters: if your money page would itself be defensible as truthful and substantiated advertising, the cloaking is a network-policy issue, not a federal-law issue. If your money page wouldn't be defensible, you're in legal trouble whether or not you cloaked.

Cases worth reading for context:

FTC v. BetterHelp (2023): FTC press release and stipulated order. $7.8M settlement. The deception was around data sharing and the framing of the service. Cloaking was not the central charge; the underlying deception was.

FTC v. Goli Nutrition (2024): FTC complaint. Native and affiliate advertising of supplement claims. Again, the deception is about substantiation; cloaking was incidental.

FTC v. Tai Lopez and Mentorbox-related entities (2022 onward): Multiple FTC and FTC-adjacent enforcement actions around income-claim and "course economy" advertising. The SEC's parallel Tai Lopez actions on his Pier 1 / Modell's / Stein Mart-related fund are also instructive context for how the FTC and SEC view post-acquisition affiliate-style advertising.

FTC's 2015 Enforcement Policy Statement on Deceptively Formatted Advertisements: the canonical native-ad framework. This is the document that establishes the standard the FTC actually applies.

FTC v. Roomster (2022): settlement announcement. Fake reviews are themselves deception; the platform-distribution layer is secondary.

The pattern across these cases: the FTC pursues advertisers and platforms for the underlying deception, not for the technical mechanism by which the deception was delivered. If you cloak a page that's truthful, the FTC has no case. If you don't cloak a page that's deceptive, you still have an FTC case.

Network policies on cloaking

Every major ad network explicitly prohibits cloaking in their advertiser policies. The relevant clauses:

Outbrain: Acceptable Use Policy prohibits "submitting an Ad ... that is materially different from the content that visitors actually see when they click on the Ad" — explicit anti-cloaking language.

Taboola: Advertising Content Policies prohibits "redirecting visitors to a different landing page based on user agent, IP address, or other characteristics" without legitimate business justification.

Google Ads: Misrepresentation policy is the most aggressive. Google has a multi-billion-row corpus of compliance signals and dedicated cloaking-detection infrastructure. They are the hardest network to cloak against and the fastest to ban.

Meta (Facebook): Advertising Standards explicitly prohibits cloaking and similar circumvention. Meta has historically been more aggressive than the native networks here.

The penalty for getting caught varies. On the native networks, single-account suspension is the typical first outcome for a single cloaked creative. Pattern-of-cloaking can result in entity-level bans where the network refuses to do business with you under any name, having identified your fingerprints.

What compliance teams actually catch

Compliance teams at the major networks use a combination of automated and manual detection:

Automated detection:

Render-vs-DOM comparison. Periodically rendering the lander URL with a real headless browser and comparing the rendered DOM to what was submitted. If the page looks different, flagged.
IP-based testing. Compliance reviewers often access landers from IP ranges the advertiser doesn't expect — residential IPs in the target geo, datacenter IPs, mobile IPs. If the page changes between these, flagged.
User-agent rotation. Same logic.
Headless-vs-headed detection. Modern compliance tooling tries both. If JavaScript-detection logic on the lander tries to differentiate, flagged.
Referrer-substitution. Hitting the lander URL directly, with the ad-network's expected referrer, with a Google Search referrer. Inconsistencies are flagged.
Time-of-day testing. Some cloakers serve the white page during business hours. Compliance reviewers run tests off-hours.

Manual detection:

Reviewer accesses lander from real user-like environment. This catches simple cloaking.
Customer complaints. A user complains about a misleading lander; reviewer investigates.
Network-level monitoring of click-quality signals. A creative whose post-click signals look pathologically different from peer creatives gets flagged for human review. This catches sophisticated cloaking that automated tests miss.

What compliance teams miss:

Sophisticated JS-fingerprint cloaking. A cloaker that detects the specific browser fingerprint of compliance bots and selectively serves the white page only to those, while serving the money page to everyone else, is genuinely hard for automated review to catch. These do still get caught — but typically by post-launch click-quality monitoring, not pre-launch creative review.
Geographic dayparting cloaking. Showing the white page in compliance-team timezones (often US business hours) and the money page in user timezones.
Cookie-state cloaking. Showing the white page on the first visit (when reviewers are clean) and the money page on subsequent visits.
CDN-edge cloaking. Different CDN edge serves different content based on edge geo-IP or some custom routing rule.

These work for a while. They eventually fail when network compliance investments catch up — and they have been catching up steadily. Cloudflare, Akamai, and the platform-internal teams have invested significantly in adversarial detection. The Akamai bot management research and the Cloudflare bot mitigation docs describe the detection patterns.

The grey zones, in detail

A few categories that affiliate operators do treat as cloaking-adjacent but that have legitimate non-deceptive uses:

Geo-gating offers that aren't legal in some jurisdictions. A US-only insurance product whose lander shows a "not available in your region" page to non-US visitors is not cloaking in the deceptive sense; it's appropriate compliance with state-by-state insurance licensing rules. The challenge is that the same mechanism (geo-IP-based content switching) gets used by both legitimate compliance and abusive cloakers, and the network's automated detection can't always distinguish. Solution: document the legitimate use to your account manager up front.

Device-based content switching for consent management. EU-CCPA-aware sites that show a cookie banner to EU visitors and a simpler page to non-EU visitors are technically serving different content based on geo. This is required by GDPR and CCPA, not optional. Networks generally do not flag this.

Affiliate link expansion vs the bare landing page. Showing different versions of a lander based on whether the user came from an affiliate link with a tracking parameter vs. came directly. Borderline. Generally tolerated if the differences are minor and the substantive content is the same.

A/B testing of multiple landers. Random assignment of incoming visitors to one of N pre-landers is not cloaking; it's standard CRO. The networks are aware and tolerate this if the variants are all compliant individually.

Click-fraud filtering. Sites that detect and route bot traffic away from monetized content — e.g., showing a stub page to visitors that fingerprint as bots. This is defensive cloaking against fraud, not against the network. Generally tolerated.

The line between these and abusive cloaking is intent. A geo-gate that exists for legal compliance is fine. A geo-gate that exists to hide the real money page from network reviewers in the network's HQ country is not.

What compliance teams actually do with their day

This is partly speculative — compliance team workflows are not publicly documented in detail by the networks — but based on conversations with people who have worked in those teams or alongside them, the daily workflow is roughly:

Queue review. Process the queue of newly-submitted creatives. Apply automated filters first. Spot-check.
Reactive review. Investigate creatives flagged by post-launch automated monitoring (click-quality anomalies, user complaints, brand-safety alerts).
Pattern review. Look for advertiser-level patterns — same advertiser submitting multiple borderline creatives, same lander pattern appearing across multiple advertiser accounts (hint of multi-account abuse), creatives with structural similarity to recently-banned campaigns.
Escalation review. Manually investigate the highest-stakes flags — typically anything in regulated verticals (health, finance, certain political) or anything that crossed thresholds for FTC-relevant claims.
AM coordination. Pass back compliance findings to account managers, who then coordinate with the advertiser.
Network-policy refinement. A small fraction of compliance team time is spent updating internal policies as new patterns of abuse emerge.

The teams are typically 5-30 people for a major network. They are well-paid, experienced, and not naive. Operators who imagine compliance reviewers as undertrained junior staff are mostly wrong.

What compliance teams cost the platforms

The platforms invest in compliance because the cost of getting it wrong is enormous. The FTC's settlements with platforms over deceptive third-party advertising are seven and eight figures. The reputational damage of being a "scam-y native widget" is the slow death of the publisher relationships, which is the slow death of the supply.

Outbrain, Taboola, and Google all reference compliance investment in their public IR materials. It is treated as a cost of doing business.

What this means for the operator

Practically, for an affiliate operator:

Don't cloak. Just don't. The math doesn't work. The expected cost of one detected cloaking incident (account loss, trust-graph damage, potential ban from the network entity) outweighs the expected benefit of slipping a borderline creative past compliance for a few weeks.

If your offer requires the page to be different from the ad, redesign the offer. A creative whose claims can't be substantiated on a fully-compliant lander is a creative that's better not to run.

Use legitimate geo and device differentiation only. Document the use to your account manager up front. Be ready to walk through your routing logic on request.

Monitor your own click-quality signals. If your post-click metrics start to diverge from peer norms, pause and investigate. Don't wait for the platform to find it first.

Run your tracking and your money page on the same domain when possible. Multi-domain redirect chains are a cloaking-adjacent pattern that gets flagged for review.

Treat compliance team interaction as a relationship, not an adversarial encounter. Operators who escalate transparently and quickly when compliance issues arise typically retain accounts; operators who try to bluff or stall typically lose them.

A short history of how cloaking became central to affiliate marketing

A bit of context on how we got here. In the early 2010s, the affiliate-marketing industry was dominated by Facebook ads, with a long tail of native and search arbitrage. Facebook's compliance was relatively permissive. Cloaking existed but wasn't operationally central — most operators simply ran on whichever creative cleared review.

From 2014-2017, Facebook's compliance tightened sharply, partly in response to a series of consumer-protection actions (the FTC's data and privacy actions against Facebook had broader knock-on effects on advertiser policy) and partly due to internal political pressure on the company. As Facebook tightened, operators discovered that cloaking — showing one page to Facebook's reviewers and another to users — let them keep running campaigns that the platform's literal policy prohibited.

This worked, briefly. Then Facebook invested heavily in cloaking detection. By 2018-2019, sophisticated cloaking on Facebook required serious engineering investment, and most operators either left the platform or moved to less-defended traffic sources.

The native networks were initially less aggressive on cloaking detection than Facebook, partly because their compliance teams were smaller and partly because the technical patterns of native advertising made certain forms of cloaking less obviously deceptive (geo-gating, for instance, has more legitimate use cases on a content-recommendation widget than on a feed ad). Outbrain and Taboola tightened progressively from 2019 onward but remained behind Facebook's detection capability.

By 2022-2023, the gap had largely closed. The major native networks were investing in detection at scale, and the search engines (Google, Microsoft) had been aggressive on this for years. The remaining "cloaking-friendly" surfaces became a long tail of smaller networks and grey-market traffic sources that legitimate operators mostly avoid.

The 2024-2026 period has been a continued tightening, with Cloudflare's bot detection, Akamai's mitigation, and platform-internal teams collaborating in increasingly sophisticated ways. The cost of running undetected cloaking has climbed by roughly an order of magnitude over five years.

What "white page" content looks like in 2026

Even though I've spent this whole piece explaining why operators shouldn't cloak, it's worth being concrete about what cloaking looked like operationally for those who did it, because the artifact is instructive about what the platforms are detecting against.

A "white page" — the version shown to compliance reviewers — typically had:

A simple landing page with no aggressive claims
An "About Us" section with company information
Privacy policy and terms-of-service links
Often, a deliberately bland call-to-action that nobody would actually click

A "money page" — shown to real users — typically had:

The actual offer with the actual claims
The actual call-to-action driving the conversion
Often, more aggressive scarcity framing, urgency timers, and the like

Detection happens when reviewers (or automated detection systems) successfully access the money-page version. The detection mechanisms include:

Hitting the URL from residential IPs in the target geo
Hitting the URL with rotating real-user user-agents and TLS fingerprints
Hitting the URL from known reviewer IPs that the cloaker doesn't recognize
Comparing the rendered DOM across these accesses
Looking for JavaScript-execution patterns that suggest cloaking logic

The arms race here has favored the platforms. Cloaking that worked in 2018 doesn't work in 2026.

The "soft cloaking" pattern that's actually fine

For balance, let me describe a pattern that's sometimes called cloaking but that's broadly tolerated by the platforms and is operationally fine.

Many affiliate landers personalize content based on traffic source — showing different copy to a user arriving from an Outbrain creative about insurance vs. a creative about supplements. The lander URL is the same; the rendered content differs by utm_source or by an intermediate redirect parameter.

This is technically content variation by request property, which is the definition of cloaking. But the variation is in service of relevance, not deception. The compliance team sees one variant; users see different variants depending on how they arrived; all variants are individually compliant; and the variation is documentable as a CRO practice.

The networks are aware of this pattern and generally don't flag it, provided that:

All variants are individually policy-compliant.
The variation is logically connected to the traffic source (i.e., relevant to the creative the user clicked).
The total surface area of the lander is still substantively the offer described in the creative.

This is the right way to think about content variation. It's also the right structural model: design every variant to be defensible on its own. If you can't, you've crossed into cloaking proper.

The macro trend

Over the last five years, the cost-to-cloak has gone up and the benefit-to-cloak has gone down. Detection infrastructure has improved. Network entity-level bans are increasingly hard to recover from. The FTC and state AGs have become more active in enforcing against affiliate-style deception. Search-feed partners have become more aggressive in scrubbing low-quality traffic.

The operators making real money in 2026 are largely not cloaking. They are running clean offers with substantiated claims and earning margin from operational excellence — better creative, better optimization, better account management. The ones still cloaking are mostly the ones running on borrowed time.

This is, broadly, a good development for the industry. It's also a story that the affiliate-marketing-course economy will not tell you, because it doesn't sell as well as "secret cloaking method to bypass Outbrain compliance."