Why do real publishers get flagged as MFA, and how do you catch the errors?
Part of MFA avoidance
The short answer
An MFA false positive is a real publisher excluded as made for advertising because one surface signal, such as a stock template, dense ads on short pages or a sudden traffic jump, matched a rule. Small and specialist sites are exposed because their ordinary traits resemble MFA traits. Catch errors by checking content evidence before an exclusion runs.
Key takeaways
- IAB UK says there is no single agreed definition of MFA and points to a combination of factors, so a rule that acts on one trait will misjudge some real publishers.
- No public study ranks which MFA signals cause the most false positives; treat signals with plain innocent explanations as review triggers, not exclusion rules.
- A 2026 study of 4,352 news articles on 51 domains found three classification services gave conflicting labels, so one tool's flag is a lead, not a verdict.
- A reversal is only finished when the exclusion entry itself is removed at the level it was set, and the change is confirmed in delivery reports.
- False positives can fall hardest on publishers with little bargaining power, which makes reversal records part of a fair practice.
Can a small publisher be wrongly flagged by a blunt MFA rule?
Yes. The IAB UK guide to identifying MFA websites (opens in a new tab) opens with the problem: "Currently, there is no single agreed upon definition for MFA websites." It lists ad placement and density, quality and relevance of content, traffic patterns, and navigation and user journeys, then says "A combination of the above factors can help identify a Made For Advertising site." A rule that fires on one factor skips the combination step, and small sites whose ordinary traits overlap with the list are exposed to it.
Definition
- MFA false positive
- A publisher that produces real content for a real readership but is classified, listed or excluded as made for advertising, because a rule or model read a surface trait as proof of an ad-first business.
The cost is easy to miss. An excluded site generates no impressions, so nothing in a delivery report shows what was lost. The publisher may never learn why its fill rate dropped, and the advertiser loses reach on exactly the specialist pages a brief asked for. The cluster hub covers the policy side of disputes; this page is about finding the error before and after it happens.
Which signals cause the most false positives?
No public study measures which MFA signals produce the most wrongful flags in the United States or Canada, so no ranking is given here. What a buyer can do is list the signals that have plain innocent explanations and treat those as triggers for a person to look, never as exclusion rules on their own. The table below pairs each signal with its innocent reading and the evidence that separates the two.
| Signal | Innocent reading | Evidence that separates them |
|---|---|---|
| Stock or generic template | A one-person site using a popular theme it cannot afford to customize | Whether the articles are specific, sourced and written for a defined reader |
| High ad density on mobile | Short reference pages where a few ad slots fill much of the screen | Ad count and refresh across long and short pages, not one page |
| Sudden traffic spike | A link from a large site, a forum thread or a seasonal search surge | Whether the spike has a referrer that explains it and then fades |
| New or changed domain | A rebrand, a move from a blog platform or a merger of two sites | Redirects from the old address and the same bylines on old work |
| Short articles | Specialist answers, product notes or glossary entries | Whether each page answers its own question completely |
| No audience data in the target market | A niche site below public popularity thresholds | Recorded as missing data, never as zero reach |
| Sensitive keywords on the page | Health, identity or crime reporting written responsibly | A reading of the full article, not the keyword match |
Why can two MFA or safety tools disagree on the same site?
Classification is a judgment, and different systems weigh evidence differently. A January 2026 study on arXiv, Inconsistencies in Classification of Online News Articles (opens in a new tab) by Smith and colleagues, compared the labels three brand safety services gave to 4,352 news articles from 51 domains. The authors found that the services "often produce conflicting classifications, with significant discrepancies between providers," and that the inconsistencies can cause "revenue losses" for publishers. The study measured safety labels, not MFA scores, but the lesson carries: disagreement between tools means at least one of them is wrong on some pages.
Griffin's analysis of brand safety in platform governance (opens in a new tab) puts the incentive plainly: "Advertisers and platforms have little to lose from overinclusive brand safety measures." The same logic applies to MFA lists. A buyer who excludes one real publisher too many sees no penalty, so errors can pile up unless someone is tasked with looking for them.
How do you check an MFA flag before excluding a site?
Check a flag in six steps before it becomes an exclusion
- Name the signal. Write down exactly which trait raised the flag and which tool or rule raised it.
- Look for a second, independent signal. An exclusion needs agreement from different areas, such as layout and traffic, not two readings of the same trait.
- Open the articles. Read several recent pieces on phone and desktop and note whether they serve a specific reader with specific information.
- Find the innocent reading. Check the table of signals above for the ordinary explanation and look for evidence of it, such as an old domain redirect.
- Separate missing data from bad data. A site the tool could not measure goes to review, never straight to exclusion.
- Record the call. Save the evidence, the reviewer's name, the date and the outcome, whichever way it goes.
What should happen after a false positive is found?
Reversing the decision on paper is half the job. The exclusion entry has to come out of each place it was copied, and the format of the entry matters. Google Ads' help page on excluding webpages and videos (opens in a new tab) explains that a top-level domain entry excludes the domain with its subdomains and pages, while a first-level subdomain entry such as www.example.com leaves other subdomains open. It also notes that "Excluding placements is different from removing them."
Closing out a false positive
- The domain is removed from the shared exclusion list and from any account, campaign or ad group list it was copied to.
- Google Ads settings are checked in both places the help pages give, Tools > Content suitability and Admin > Account settings, since the pages disagreed on the path when checked in October 2026.
- In Display & Video 360, the line item's Apps and URLs targeting is edited and rechecked after a few hours, since Google says bidding updates can lag.
- The site's record shows the original flag, the evidence that overturned it and the reviewer.
- The rule or threshold that raised the flag is reviewed, so the same pattern does not catch the next specialist site.
- The publisher is told the outcome if it raised the dispute.
The Display & Video 360 help page (opens in a new tab) says that after a targeting change on an active line item "a few hours may pass before Display & Video 360 updates your line item's bidding accordingly." Confirm the reversal in the next delivery or placement report rather than at the moment of the edit. For where list entries live and how they are structured, see our guide to the MFA exclusion list.
How do MFA false positives affect fairness?
Errors do not land evenly. Griffin writes that "content classification tools are also inevitably prone to errors and bias," and that over-inclusive measures produce "a systematic loss of resources and visibility for marginalised creators." The same analysis names the revenue of publishers covering serious news and topics relevant to marginalized groups as one casualty. A small specialist site has no account team to call and no leverage to ask why its demand fell.
There is also a market cost. In a quasi-experiment at a European ad exchange, Ada, Abou Nabout and McDonnell Feit found that sharing placement information raised average revenue per impression, and they recommend it for "smaller volume, high quality sites" (arXiv, December 2020 (opens in a new tab)). Small volume and high quality can coexist, which is the combination a blunt MFA rule can misread. Our guide to fair publisher planning covers measuring outcomes by publisher size.
What does a false positive look like on a real brief?
Fictional example
A film camera and darkroom brief in Alaska
A fictional Anchorage shop selling film cameras, darkroom chemicals and camera repair asks its agency for a display plan in Alaska, around Anchorage, Fairbanks and Juneau, beside analog photography guides, camera reviews and local arts coverage. The agency's MFA rule set flags three of the eleven candidate sites.
A Fairbanks film developing blog is flagged for a stock theme and four ad slots on short pages. The reviewer opens six posts: each documents a developing test with the writer's own scans, chemical ratios and results. The second signal the policy requires is missing, so the flag is overturned and the rule's weight on templates goes to the quarterly review.
A camera repair guide site is flagged for a traffic spike. The referrer data the publisher shares shows a link from a large photography forum, and the spike fades within a week, so it is cleared. The third site, a gallery of camera listicles with rotating ad slots on each slide and paid traffic links, matches several signals across areas and stays excluded, with the evidence attached.
How does PlacementGraph limit wrongful MFA exclusions?
Frequently asked questions
Can a site be a false positive one quarter and a real MFA site the next?
Yes. Ownership changes, new traffic sources and new ad layouts can turn a cleared publisher into an ad-first business. A reversal clears the evidence that existed on the review date, so keep the date on the record and let the next scheduled check, or a trigger such as a new owner, reopen it.
Who should be able to overturn an MFA flag?
Someone other than the person or tool that raised it, with authority over the exclusion list. A second reviewer catches the anchoring that comes from seeing the flag first. The reviewer's name and evidence go on the record, so a later audit can see why the site came back.
Does a false positive cost the advertiser anything?
It costs reach and relevance. The excluded site's readers can be the specific audience a brief describes, and the budget moves to whatever inventory is left. Because an excluded site serves nothing, delivery reports never show the loss, which is why reversal counts are worth tracking.
Should a publisher be told why it was flagged?
Telling a publisher which signal raised a flag lets it supply the missing evidence, such as referrer data for a traffic spike or redirects from an old domain. The detail of any scoring model can stay private; the trait and the evidence the reviewer needs are enough for a useful reply.
Sources
- A guide to identifying Made For Advertising websites (opens in a new tab)IAB UK, July 2023
- Inconsistencies in Classification of Online News Articles: A Call for Common Standards in Brand Safety Services (opens in a new tab)Michael Smith, Riley Grossman, Antonio Torres-Aguero, Pritam Sen, Cristian Borcea and Yi Chen, arXiv (January 3, 2026)
- From brand safety to suitability: advertisers in platform governance (opens in a new tab)Rachel Griffin, Internet Policy Review 12(3) (July 11, 2023)
- Context information increases revenue in ad auctions: Evidence from a policy change (opens in a new tab)Sıla Ada, Nadia Abou Nabout and Elea McDonnell Feit, arXiv (December 1, 2020)
- Exclude specific webpages and videos (opens in a new tab)Google Ads Help
- App, channel, and website targeting (opens in a new tab)Display & Video 360 Help
- About content suitability (opens in a new tab)Google Ads Help
- Set content exclusions for Display campaigns (opens in a new tab)Google Ads Help
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.