How do you run a YouTube brand safety audit after a campaign?
Part of YouTube brand safety
The short answer
A YouTube brand safety audit checks where a campaign's ads actually served against the brand's written rules. The buyer exports served channels and videos from Google Ads reports, reviews a documented sample, logs each finding with evidence, turns confirmed problems into exclusions or setting changes, and reports what was covered and what the reports could not show.
Key takeaways
- An audit starts from served placements, not from the plan, because ads can reach videos nobody chose in advance.
- Google says the Where ads showed report is not an exhaustive list of placements and is not a billing record, so the audit report must state those limits.
- If a random sample of N videos shows no problems, the upper limit of the 95 percent confidence interval for the problem rate is about 3/N, which gives a defensible sample size.
- Each finding needs a severity, a cause and an action, and should say whether the brand's controls could have caught it.
- TAG's certification guidelines ask participating buyers to review inclusion and exclusion lists quarterly, a useful floor.
What does a post campaign YouTube audit check that vetting does not?
Vetting judges channels before money moves; an audit judges delivery after it has moved. The two answer different questions. Pre launch review, covered in the channel vetting guide, asks whether a channel deserves a place in the plan. A post campaign audit asks where the ads really ran, whether those videos met the brand's rules, and whether the account's settings did what everyone assumed. A campaign that ran on topics, keywords or audiences rather than a fixed channel list can reach thousands of videos that no planner looked at.
Definition
- YouTube brand safety audit
- A dated review of the channels and videos where a campaign's ads served, compared with the brand's written suitability rules, producing a findings log, exclusions or setting changes, and a report that states its own coverage limits.
How do you audit where YouTube ads actually ran?
Start with Google's own delivery records. The Where ads showed report (opens in a new tab) lists channels, videos and websites at the account, campaign or ad group level, under Insights and reports. It opens at channel level; selecting channels and choosing View Details shows video level rows as Placement (Group) and Placement (Detail). The Content suitability report (opens in a new tab) adds video detail through the Content suitability Placement name (details) column and, for Shorts, shows "placement data for content that was shown before and after the YouTube Shorts ads appeared."
Seven steps to pull and prepare the served placement data
- Fix the audit window. Use the flight dates, or the last full month for an always on campaign, and write them at the top of the log.
- Export at channel level first. Sort by impressions so the channels that carried the most delivery are visible.
- Drill down to videos. Use View Details for the channels in scope, since a brand rule is broken by a video, not by a channel name.
- Separate the surfaces. Keep watch page rows, Shorts adjacency rows and rows reported only as youtube.com apart, because Google reports feed placements as youtube.com.
- Mark deleted videos. Rows labeled Unknown Channel or not available still carry impressions and go in the log as unreviewable.
- Record the settings in force. Note the inventory type, excluded themes, keywords and placement lists active during the window.
- Freeze the export. Save a dated copy, since Google says impression counts in these reports should not be treated as final.
How many served videos should a YouTube audit sample?
No platform or industry body sets a number, so the sample size should be justified rather than borrowed. In a 2022 paper on sample size justification (opens in a new tab), Daniël Lakens lists six approaches, among them "collecting data from (almost) the entire population," "choosing a sample size based on resource constraints" and "planning for a desired accuracy." An audit can use all three: review the whole population where it is small, state the hours available, and say what precision the sample buys.
For the precision part, a long standing rule of thumb helps. As a 1995 JAMA letter (opens in a new tab) summarizes it, "if 0 events are observed in N trials, the upper limit of the 95% confidence interval (CI) for the frequency of events is approximately 3/N." Applied to an audit, a random sample with zero breaches still leaves room for a breach rate up to about 3 divided by the sample size.
| Videos sampled at random, zero breaches found | Breach rate still not ruled out | Fits |
|---|---|---|
| 30 | About 10 percent | A quick check on a small flight |
| 60 | About 5 percent | A standard monthly audit |
| 100 | About 3 percent | A client with a strict rule set |
| 300 | About 1 percent | A sensitive category or a dispute |
Combine this with a census of the head. Review all videos in the channels that carried the largest share of impressions, then draw the random sample from the remaining tail, and state both parts. Review Shorts and watch page placements as separate samples, since Google reports them differently and the brand may hold them to different rules.
Which YouTube audit findings should trigger exclusions?
Classify each finding against the brand's rules first and against YouTube's advertiser friendly content guidelines (opens in a new tab) second. Google's content suitability page (opens in a new tab) explains why both are needed: "Our policies define the type of content that creators and publishers are allowed to monetize," while suitability settings "provide additional filtering to help you find the right fit for your product or brand." A video can pass the first and fail the second.
| Finding | Severity | Action |
|---|---|---|
| A served video breaks a refusal the brand set in writing | High | Exclude the video by ID; review the channel's other served videos |
| Several served videos on one channel break the same rule | High | Exclude the channel and record the pattern |
| A served video is near the line, such as tense news commentary | Medium | Log it; decide with the client whether the rule needs wording |
| Breaches cluster in one theme the account did not exclude | Medium | Change the account setting rather than adding single exclusions |
| Video deleted or private before review | Unknown | Log as unreviewable with impressions; no exclusion |
| Served video fits the brief and the rules | None | No action; keep it as evidence the controls worked |
Before logging a finding as a control failure
- The surface was covered: Google's coverage table shows the control applies to that surface, for example placements but not keywords on Shorts.
- The campaign type was covered: Google states YouTube Sponsorship campaigns bypass account level content suitability exclusions.
- The exclusion existed during the window and had taken effect; Google says new placement exclusions typically take effect within 12 hours.
- The video was judged as it looked when the ad served, not after a later re-edit, where that can be established.
How should YouTube audit results be reported to a brand?
Report in the brand's terms and show the method before the findings. The client needs to know what was reviewed, what could not be, what was found and what changed as a result. Keep the full log as an appendix and lead with a one page summary.
| Column | What it holds |
|---|---|
| Placement | Channel name and ID, video ID, surface (watch page, Shorts, youtube.com only) |
| Delivery | Impressions in the window, from the frozen export |
| Rule | The brand rule or YouTube guideline it was judged against, quoted |
| Evidence | What the reviewer saw, with the review date |
| Severity and cause | High, medium, unknown or none; control gap, missing rule or platform limit |
| Action | Video exclusion, channel exclusion, setting change or none, with level and date |
| Verified | Date the change was confirmed in the account's change history |
Google's placement exclusions page (opens in a new tab) points to Change history to confirm account level exclusions, which gives the Verified column its proof. A Display & Video 360 buyer has a parallel record: Google's brand controls page (opens in a new tab) describes a view of "active settings across all of your line items and YouTube ad groups" and a downloadable list of changes.
How often should a YouTube brand safety audit run?
Set a calendar and add triggers. For a floor, the TAG Brand Safety Certified Guidelines (opens in a new tab) (version 5.0, July 2026) require participating direct buyers and intermediaries to document how they maintain inclusion and exclusion lists and to execute those procedures "as part of their internal quarterly review process." That requirement binds companies in the program, not all advertisers, but quarterly is a defensible minimum for any always on YouTube buy.
- At the end of a flight, before the client report.
- Monthly for always on campaigns that use topics, keywords or audiences rather than a fixed channel list.
- Within days of a sensitive event or a public complaint, scoped to the affected window.
- After a major account change, such as a new inventory type or a campaign type upgrade.
What does a YouTube audit look like on a real campaign?
Fictional example
A pet insurance provider in Oklahoma
A fictional pet insurance provider runs a two month YouTube campaign for dog and cat owners in Tulsa and Oklahoma City, on topics for pets and animal care. Its rules refuse graphic injury content and content that mocks animal suffering.
At the end of the flight, the agency exports served placements, reviews all videos on the ten channels with the largest delivery, and draws a random sample of 100 videos from the remaining tail. Shorts adjacency rows are sampled separately. The method and the 3/N limit go on the first page of the report.
The sample finds a cluster of veterinary surgery videos with graphic footage on two channels, which the account's settings did not exclude. The agency excludes both channels, proposes a narrower inventory type for the next flight, logs four deleted videos as unreviewable, and confirms the exclusions in change history the following day.
How does PlacementGraph help after a YouTube audit?
Frequently asked questions
Does the Where ads showed report match the invoice?
Not necessarily. Google says the report is not an accurate record of billing, that impression counts are not final, and that billing invoices are the source of truth for billable impressions. An audit should cite the frozen export for delivery shares and leave spend reconciliation to the invoice.
Can an audit review ads that ran in the YouTube Home feed?
Only partly. Google says ads shown outside the watch page, such as the Home feed, appear in the Where ads showed report as youtube.com rather than as channels or videos. The audit can report their share of delivery but cannot review the surrounding content, and the report should say so.
What happens to a flagged video that was deleted before review?
Google's reports keep its impressions under Unknown Channel or a not available label. Log it as unreviewable with its delivery, and exclude nothing on its account. If several deleted videos sit on one channel, review the channel's remaining uploads before deciding.
Should YouTube Sponsorship campaigns be audited the same way?
Audit them against their own lineup. Google states that YouTube Sponsorship campaigns automatically run on Moderate inventory and bypass all account level content suitability exclusions, so a breach there is not a failure of account settings. The finding belongs with the sponsorship's channel selection.
Sources
- About the Where ads showed report (opens in a new tab)Google Ads Help
- About the Content suitability report (opens in a new tab)Google Ads Help
- About content suitability (opens in a new tab)Google Ads Help
- About placement exclusions (opens in a new tab)Google Ads Help
- View and edit brand controls for your advertiser (opens in a new tab)Display & Video 360 Help
- Advertiser-friendly content guidelines (opens in a new tab)YouTube Help
- Brand Safety Certified Guidelines, version 5.0 (July 2026) (opens in a new tab)Trustworthy Accountability Group (TAG)
- Sample Size Justification (Collabra: Psychology, 2022) (opens in a new tab)Daniël Lakens, Eindhoven University of Technology
- If Almost Nothing Goes Wrong, Is Almost Everything All Right? (JAMA, 1995) (opens in a new tab)JAMA Network
Researched and written by PlacementGraph Research. Read how we research, source and update these pages. Corrections: diana@placementgraph.com.