Google Ads Personalized Advertising Policy 2026: Which Audiences You Cannot Build, and What Changes on October 30
Three symptoms, one root cause. A remarketing list sits in Audience Manager but refuses to attach to a campaign. A Customer Match upload matches at 68% and still cannot be applied. Demographic settings quietly disappear from a campaign you have been editing for months. None of these produce a disapproval notice, so most advertisers spend days auditing bids and budgets. The actual cause is the personalized advertising policy — a separate rulebook that sits on top of regular ad policy and enforces itself silently.
This guide covers how that rulebook works in 2026: which interest categories lock you out of first-party audiences entirely, where demographic and geographic targeting is restricted by law rather than by product, which data practices put the account itself at risk, and exactly what changes on October 30, 2026. It ends with a pre-launch checklist worth running before the first dollar goes through the account.
What the personalized advertising policy actually covers
Personalized advertising is any delivery decision that uses data about the person rather than the context of the moment. That is a wider net than most buyers assume:
- remarketing lists and similar segments;
- affinity, in-market and custom intent audiences;
- Customer Match and any uploaded first-party list;
- demographics — age, gender, parental status, household income;
- audience signals in Performance Max and Demand Gen;
- dynamic remarketing driven by a product feed.
The crucial distinction: the personalized advertising policy does not regulate what you sell. It regulates the attribute you use to pick people. A perfectly compliant offer, a clean landing page and a verified account will still hit the wall if the audience is built on an attribute Google treats as sensitive.
The second distinction is how enforcement feels. Break regular ad policy and you get a disapproval with a reason attached. Break the personalization rules and a capability simply stops existing — the segment is marked ineligible, the audience refuses to apply, impressions redistribute. There is often no notification at all.
If an audience stops working and change history is clean, do not start with bids. Start with the attribute the segment is built on and the category Google has assigned to your business.
Sensitive interest categories: the hard stop
Google maintains a set of sensitive interest categories. The underlying logic is straightforward: these are attributes where the mere fact that an ad was shown reveals something the person never chose to disclose. Advertisers whose product falls into one of these categories cannot use advertiser-curated audiences at all — no remarketing lists, no Customer Match, no similar segments.
| Group | What it covers | Where it bites in practice |
|---|---|---|
| Health | conditions, diagnoses, treatments, mental health, disability, contraception, reproductive health | supplements, clinics, telehealth, pharmacy offers |
| Hardship and trauma | abuse, violence, addiction, difficult life events | rehab facilities, injury law firms |
| Financial status | debt, delinquency, hardship, credit standing | payday lending, debt relief, refinancing |
| Identity and belief | race, ethnicity, national origin, religion, political affiliation | faith-based and ethnic retail, political projects |
| Sexual orientation and gender identity | orientation, transgender identification, intimate interests | dating, community media |
| Regulated categories | gambling, alcohol, prescription drugs | iGaming, betting, alcohol ecommerce |
The mechanism matters. The restriction attaches to the advertiser’s category, not to how you labelled the list. Renaming “visitors to the diabetes page” as “all site visitors” changes nothing — the system reads the product and the landing page. Nor does building the segment in GA4 and importing it help: the block applies at the moment the segment is attached to a campaign, regardless of where it was created. For the underlying mechanics of segment eligibility, see our breakdown of Audience Manager thresholds and membership duration.
What you keep
Losing first-party audiences is not the same as losing the channel. What remains fully available:
- contextual targeting — keywords, topics, placements, search themes in Performance Max;
- geographic targeting at city, region and country level;
- ad scheduling and device bid adjustments;
- optimized targeting, within whatever the system infers on its own — but without your lists feeding it.
In practice this shifts the center of gravity from audience engineering to intent and landing page quality. We covered how to get performance out of contextual signals alone in the guide to optimized targeting and audience signals.
Housing, employment and credit: a separate regime
In the US and Canada a second layer applies — the equal access categories. If your product is housing, employment or consumer finance, the campaign must be flagged as such, and a block of settings then disappears from the interface.
| Category | Prohibited | Still permitted |
|---|---|---|
| Housing — rentals, sales, mortgages, home insurance | targeting or excluding by gender, age, parental or marital status; ZIP code targeting | city, region, country; radius of 1 km or larger |
| Employment — job listings, recruiting, career services | same | same |
| Consumer finance — loans, cards, refinancing | same | same |
Two details catch people out. First, the restriction covers exclusions as much as inclusions — “do not show to people over 55” is the same violation as targeting them. Second, radius targeting below 1 km is gone, which breaks any micro-geo strategy built for local lead generation. For the general mechanics of demographic layers, see demographic targeting and audience exclusions in Google Ads.
What changes on October 30, 2026
September’s policy update takes effect on October 30, 2026, and for once it widens rather than narrows. On YouTube inventory, personalization is now permitted for alcohol, alcohol-related products and alcohol alternative beverages, where local law allows it.
- alcohol brands gain access to audience tooling on YouTube that was previously closed to them;
- age restrictions remain in full force — personalized ads to minors stay prohibited;
- targeting on alcohol-related health attributes stays prohibited; the Health category does not open;
- the change does not apply in Egypt, India, Indonesia or Poland;
- licensing, local law and platform age-gating requirements are unchanged.
Worth stating plainly: this is YouTube inventory only, and alcohol only. Gambling, prescription drugs and the remaining regulated categories are untouched. If you are planning a November YouTube push, check where your ads can actually appear first — see brand suitability and inventory modes in Google Ads.
Data rules: three mistakes that cost accounts
The second half of the policy governs the data itself, not the targeting. These mistakes are the expensive ones, because they lead to suspension rather than a disabled feature.
1. PII in ads and feeds
You may not collect personally identifiable information inside ad formats unless the format was designed by Google for that purpose. A Google lead form qualifies. A custom form embedded in a creative, an email in a URL parameter, a phone number in a UTM tag does not. You also may not pass PII through data feeds, or mix PII with anonymous identifiers inside a single segment.
Hashing is the common misunderstanding. Hashing an email before a Customer Match upload makes that specific flow compliant, with its own consent requirements. It is not a universal pass — the same hash dropped into a tracking parameter is still a violation. For the compliant path, see Data Manager and first-party data in Google Ads.
2. Audiences that are too narrow
Google’s own wording prohibits using data segments “in a way that targets an overly narrow or specific audience.” This is an anti-deanonymization rule: the segment must not be constructible such that the ad delivery itself identifies one person. In practice the violations look like stacked constraints — one page visit, plus a tight radius, plus a single device type, plus a narrow age band.
Minimum audience size thresholds exist as a technical guardrail, but clearing a threshold is not a defense if the intent of the construction was to reach an individual.
3. Negative lists used for positive personalization
Explicitly prohibited: using exclusion lists to achieve positive personalization. The pattern this closes is building a segment on a sensitive attribute, never targeting it directly, and instead excluding everyone else — arriving at exactly the targeting the policy forbids. This reads as deliberate circumvention, and it is treated accordingly.
How a violation actually surfaces
| Symptom | Likely cause | Check first |
|---|---|---|
| Segment exists but will not attach | advertiser category classified as sensitive | landing page subject matter and declared business category |
| Demographics missing from campaign settings | campaign flagged as housing / employment / credit | campaign type and account-level settings |
| Impressions drop with no disapprovals | part of the audience stack became ineligible | segment statuses in Audience Manager |
| Ads serve to an unusually small pool | limited ad serving | our guide to limited ad serving |
| Account suspended without warning | a data violation — PII or circumvention | what flows into feeds, forms and URL parameters |
One more confound: if consent signals are not passed correctly, segments simply fail to populate, which looks identical to a policy restriction. Separating the two starts with the measurement layer — see Consent Mode v2 and advertising privacy and the broader context in third-party cookies and Privacy Sandbox.
Three strategies that work in a restricted category
Losing audience tooling changes how you steer, not how much traffic you can buy. Three approaches hold up.
Intent instead of audience
Move all segmentation into query structure. Rather than “show this to people who visited,” the unit becomes “show this to people phrasing the query the way bottom-of-funnel buyers phrase it.” Practically: informational queries in one campaign with a modest bid and their own landing page, transactional queries in another campaign carrying the budget. The quality of that structure is entirely a function of negative keyword discipline, which turns keyword research and query mining from hygiene into the primary steering mechanism.
Placements instead of remarketing
For video and display, placements and topics absorb the role remarketing normally plays. Hand-picked placements produce a more predictable audience than any automatic segment — at the cost of labor. Run two lists in parallel: a whitelist of placements that convert and a blacklist of placements that spend, and revisit both weekly. Mechanics in placement exclusions and brand safety.
Separate accounts for separate products
The most common structural mistake is housing a sensitive and a non-sensitive product in one account. Category is assigned at the advertiser level, so the restriction swallows both lines — you lose audiences where they would have been entirely permitted. The correct architecture is one account per line, with its own domain, landing pages and conversion set. It also removes cross-contamination of learning data between unrelated campaigns. See Google Ads account structure for scale.
One more thing worth watching
Policy classification is not static. A product that sat outside the sensitive categories can move inside them after a landing page rewrite, a new claim in the copy, or an expansion into an adjacent offer. Nobody notifies you when that happens — the audiences simply stop working one day. Treat category classification as something to re-check whenever the offer or the landing page materially changes, not as a one-time decision made at launch.
Pre-launch checklist
- Classify the product honestly — by what the visitor sees on the landing page, not by how the creative is worded. If the page is about a health condition, financial hardship or identity, do not plan an audience-led strategy.
- Check geography. Equal-access rules apply in the US and Canada; the sensitive category rules apply everywhere.
- Audit your parameters. Walk every tracking template, feed and form. No emails, phone numbers or names in URLs.
- Inspect your exclusion logic. If the exclusions sum to sensitive targeting, rebuild them.
- Design the fallback first. In restricted categories, intent-and-context architecture has to be the plan, not the recovery.
- Diary October 30. If alcohol on YouTube is relevant, verify audience tooling availability after that date and confirm your markets are not on the exclusion list.
If the account is launching into a category where half the toolkit is unavailable, the architecture should be settled before the first deposit. That is part of how we set up Google Ads agency accounts — campaign structure and available targeting types get decided at the start. Where the product requires business confirmation, see Google Ads account verification.
Related reading
Policy is one of three ceilings that shape what you can build in an account. The other two are technical: Google Ads account limits, which cap campaigns, targeting items and assets, and the rules for using someone else’s content in creatives — covered in creator partnerships boost and YouTube creator videos in campaigns.
FAQ
Is the personalized advertising policy the same as regular Google Ads policy?
No. Regular policy governs what you may advertise and how the ad must look. The personalized advertising policy governs the attributes you may use to select an audience. Full compliance with one says nothing about the other.
My audiences were restricted — is that permanent?
The restriction is tied to advertiser category, not to a time period. While the product sits in a sensitive category, first-party audiences stay closed. If the account carries multiple product lines and only one is sensitive, split them into separate accounts rather than trying to coexist.
Can I build the audience in GA4 and import it to get around the block?
No. Enforcement happens when the segment is attached to a campaign, regardless of where it was built — Google Ads, GA4 or a Customer Match upload.
What counts as PII here?
Anything that identifies an individual: name, email, phone, precise address, document numbers. Hashing does not remove data from scope; it only legalizes specific flows such as Customer Match.
Is remarketing banned outright for supplements and wellness?
If the product is positioned around a health condition, first-party audiences are unavailable. A general wellness product with no condition claim may fall outside the category — but the landing page content, not the ad copy, determines that.
What exactly changes on October 30, 2026?
Personalization opens for alcohol and alcohol alternative beverages on YouTube inventory where locally permitted. Egypt, India, Indonesia and Poland are excluded. Age restrictions and the ban on health-attribute targeting remain.
Do the housing, employment and credit rules apply outside the US and Canada?
The equal-access regime is US and Canada. The sensitive category rules and the prohibition on overly narrow audiences apply globally.
Why is radius targeting under 1 km prohibited in those categories?
A micro-radius combined with other attributes can isolate a single building or a handful of households — effectively targeting by place of residence, which in these categories maps to discrimination on a protected attribute.
How do I tell a policy restriction from a budget or bidding problem?
Check segment status in Audience Manager and whether demographic controls are still present in the campaign. Ineligible segments or missing demographic settings point to policy. If both look normal, go to change history and campaign diagnostics instead.
Can the restriction be appealed?
An appeal makes sense when the category assignment is wrong — a product classified as medical that is not. If the classification is correct, there is nothing to appeal: it is a rule, not a penalty.
Does this affect Performance Max?
Yes. Audience signals in PMax are the same segments, so they are unavailable in sensitive categories. The campaign will still run, steering on contextual signals and search themes alone.
Which is worse — a targeting violation or a data violation?
A data violation. Targeting violations usually disable a capability; data violations, particularly ones that look like deliberate circumvention, put the account at risk of suspension.
