What should happen to a site with borderline MFA evidence?
Part of MFA avoidance
The short answer
A borderline MFA site was checked, but its evidence points both ways. It should neither run nor be blocked by default. Put it in a review queue with a named owner and a deadline, have a person reread the pages behind the mixed signals, then hold it, approve a limited test, exclude it or clear it, and record why.
Key takeaways
- Borderline is a result, not a gap: the site was read and the signals disagree, unlike a site that could not be checked.
- Four outcomes cover the queue: hold, limited test, exclude and clear, each with a written condition for leaving it.
- No industry body sets a deadline for an individual site, so the review time is an operating rule the agency writes and the client approves.
- A small test budget limits spend but is not a hard ceiling, and good delivery numbers do not prove a site is not made for advertising.
- The decision record names the signals, pages read, reviewer, outcome and date, so the call can be checked or reversed.
What should happen to a site with inconclusive MFA evidence?
Inconclusive evidence means the review ran and produced a split result. Some sampled articles carry a heavy ad load while others read cleanly, or the writing is original but a new section looks off topic. The trade groups behind the September 2023 MFA definition (opens in a new tab) list five characteristics that appear in combination and leave the line to each advertiser, so a single mixed reading does not settle the question either way.
Definition
- Borderline MFA site
- A candidate site whose made-for-advertising review finished but returned signals that point in different directions, so the result is too close for an automatic keep or an automatic exclusion.
Keep borderline sites apart from two neighbors. A site that could not be read belongs with unverifiable sites, and a placement no classifier has rated belongs with unclassified inventory. The MFA avoidance hub sets the thresholds; this page covers one site after it lands in the middle.
| Outcome | When it fits | What runs meanwhile | How it ends |
|---|---|---|---|
| Hold | The deciding pages need a second look or fresh samples | Nothing; the site stays off live lines | A reviewer clears it, excludes it or moves it to a test before the hold expires |
| Limited test | The editorial work holds up but one layout or traffic clue is unresolved | A separate, capped line limited to the site, set up after approval | The reviewer reads delivery evidence and decides keep or exclude |
| Exclude | A second review confirms the clues across sections, not on one page | Nothing; the domain goes on the client's exclusion list with evidence | Only a dispute or a later review with new evidence reopens it |
| Clear | The clue turns out to be isolated or explained | The site joins the inclusion list with its reason | Normal review rhythm resumes |
Which page quality clues require a person to review a site?
A clue needs a person when it can be read two ways and the reading depends on context a rule cannot see. Google's guidance on helpful content (opens in a new tab), updated October 5, 2026, is written for site owners, but its questions make useful reviewer prompts. It asks whether content provides "original information, reporting, research, or analysis" and whether a site has "a primary purpose or focus," and it lists as a warning sign "changing the date of pages to make them seem fresh when the content has not substantially changed."
| Clue in the evidence | Innocent reading | Concerning reading | What the reviewer checks |
|---|---|---|---|
| Dense ads on some sampled articles only | One template is used for long guides with more slots | Ad load scales with traffic bought for those pages | Whether dense pages are the ones linked from sponsored units elsewhere |
| A new section far from the site's subject | The publisher is expanding coverage | Content on unrelated topics produced to catch search traffic | Bylines, sourcing and whether the section links back into the core site |
| Recent dates on old articles | Real updates with new facts | Dates changed without substantial edits | An archived copy of the same article for changed text |
| Sponsored link units on the pages | A routine recirculation or ad partner | Units that send readers straight to sister sites | Where the links lead and who owns the destinations |
Reviewer steps for one borderline site
- Open the flagged pages first. Start with the exact articles that produced the mixed signals, not the homepage.
- Add fresh samples from other sections. Read recent articles from at least two sections the first pass did not cover, on a phone and on a desktop.
- Separate the clues by area. Sort them into layout, content and traffic, because agreement across areas weighs more than repetition within one.
- Look for the innocent explanation. Write down the most likely ordinary reason for each clue and what evidence would rule it out.
- Name the outcome and the condition. Choose hold, test, exclude or clear, and state what would change the call.
How long should a borderline site wait for review?
No standards body publishes a deadline for resolving an individual site, so the wait is an operating rule the agency proposes and the client approves. The closest published cadence covers whole lists: TAG's Brand Safety Certified Guidelines (opens in a new tab), version 5.0 of July 2026, has a section titled "Ensure Inclusion/Exclusion Lists Are Reviewed Quarterly" for direct buyers and intermediaries, and asks them to document how properties are added to and removed from those lists.
Can a borderline site run with a small budget?
It can, when the client approves the test and the test asks a question delivery can answer, such as which pages actually serve the ads and what the layout looks like on those pages. In Display & Video 360 a test line can be built in advance and saved as Draft, which Google describes as the status to use "if you don't want the line item to be eligible to serve once you've saved it," according to its line item setup guide (opens in a new tab). The reviewer can prepare the line before approval.
A small budget is a limit, not a guarantee. Google's budget and pacing help (opens in a new tab) warns that "the smaller the budget and the greater the inventory available to spend on, the higher the chance of overpacing and rapidly overspending," and that advertisers remain responsible for actual cost when it surpasses the budget caps they set. A test also cannot prove a site is clean on delivery metrics alone, because the ANA's definition notes that MFA sites show high measurability, good viewability and low invalid traffic.
Conditions for a limited test on a borderline site
- The client approved the test in writing, with the site, the cap and the end date named.
- The test runs on its own line, targeted to the one site, so its delivery is easy to isolate.
- Pacing is set to spread the budget across the flight rather than to spend as fast as possible.
- The reviewer has listed the questions the test answers, such as served pages and observed layout.
- Click and viewability rates are not the deciding evidence.
- The test has an end date, after which the site is kept or excluded, never left running by default.
How do you record the final decision on a borderline site?
The record lets someone else check the call months later without the reviewer in the room. TAG's guidelines ask certified buyers to document how properties are added to and removed from their lists, and a per site record is the evidence behind that method.
| Field | What it holds |
|---|---|
| Site and list version | The domain and the version of the list it was reviewed for |
| Signals that triggered review | Each clue, its area (layout, content, traffic) and the source of the reading |
| Pages read | The URLs sampled in each pass, with the date and device |
| Innocent explanations tested | What was checked to rule each one in or out |
| Outcome and condition | Hold, test, exclude or clear, and what would reopen the call |
| Reviewer and approver | Who made the call and who approved a test or exclusion |
| Dates | Flag date, decision date and the next scheduled recheck |
Mirror the outcome in the buying platform. Display & Video 360's event history (opens in a new tab) records changes and lets users "Add a note" to a change description, and a download gives "a CSV file of the changes in your date range." A note that points to the decision record ties the setting change to its reason. Allow for lag: Google's targeting help (opens in a new tab) says "a few hours may pass" before bidding reflects a targeting update on an active line, and Google Ads says new account level placement exclusions (opens in a new tab) "typically take effect within 12 hours."
What does a borderline review look like on a real brief?
Fictional example
A medical scrubs retailer in Ohio
A fictional Columbus retailer of medical scrubs and nursing shoes asks for an English display plan in Columbus, Cleveland and Cincinnati, beside nursing career advice, hospital workplace news and nursing school study guides. Four sites come back borderline.
A nursing exam prep blog shows dense ads on its practice question pages but clean article layouts. The dense pages are the ones a sponsored link unit on another domain points to, the clue repeats in a second section, and the site is excluded with both findings recorded.
A career advice site has original interviews with nurses but a new section of celebrity salary lists. It goes on hold; the section turns out to sit on a separate subdomain, so the main site is cleared and the subdomain excluded.
A Cleveland hospital workplace newsletter shows recent dates on older posts. An archived copy shows real updates with new figures, so the clue is recorded as explained. A Cincinnati study guide site with an unresolved ad refresh reading gets a client approved test on its own capped line for two weeks, with served pages as the question.
How does PlacementGraph route borderline sites to a person?
Frequently asked questions
What should happen when two reviewers disagree about a borderline site?
Escalate to the practice owner rather than averaging the views. Each reviewer writes down which clue decided their call and what evidence would change it. The owner rereads those pages and decides, and the record keeps both opinions, which helps later reviews see where the policy wording was unclear.
Does a borderline outcome for one client apply to other clients?
The evidence can be shared, but the outcome may differ. The ANA's guidance leaves each advertiser to set its own tolerance for MFA inventory, so a site one brand excludes may be acceptable to another. Share the decision record across client teams and let each client's approved policy set the outcome.
What reopens a borderline site after it has been cleared?
Write the trigger into the decision record when the site is cleared. Useful triggers include the same clue reappearing in a later sample, a new section on an unrelated subject, a change of owner or template, or a served page that does not match what the reviewer read. Any of these sends the site back to the queue with the earlier record attached.
Can a borderline site be added to a curated deal while it is held?
Keep it out until the hold ends. A deal's domain list is only as reviewed as its entries, and in Display & Video 360 advertisers can choose to bypass advertiser level brand suitability settings for selected deals, so a held site inside a deal can serve under looser settings than the open auction line.
Sources
- Leading Trade Groups Define “Made For Advertising” Websites (opens in a new tab)ANA (Association of National Advertisers), September 26, 2023
- Creating helpful, reliable, people-first content (opens in a new tab)Google Search Central, updated October 5, 2026
- Brand Safety Certified Guidelines, version 5.0 (opens in a new tab)TAG (Trustworthy Accountability Group), July 2026
- Create a line item (opens in a new tab)Display & Video 360 Help
- Set budgets and control your pacing (opens in a new tab)Display & Video 360 Help
- Event history (opens in a new tab)Display & Video 360 Help
- App, channel, and website targeting (opens in a new tab)Display & Video 360 Help
- Brand suitability (opens in a new tab)Display & Video 360 Help
- Exclude placements at the account level (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.