Performance Max Asset Optimization: Audit the Toggles Now

Google Ads has added two Performance Max asset optimization features that pull images and video straight from an advertiser's own website, and one of them is reported to arrive switched on. Philippine agencies now carry an Ad Standards Council duty to disclose AI-made ad content, which turns that account setting into a compliance control. Audit it this week.

Every client-services team runs on a sentence that sounds like a guarantee. Nothing goes live under the client's brand until someone on our side signs it off. That sentence still describes your workflow accurately. It no longer describes your client's Google Ads account.

Google has added two features to Performance Max asset optimization that let the platform assemble creative from an advertiser's existing web content, and the advertiser who surfaced them reports that one arrived already enabled. Five weeks earlier, the Ad Standards Council put an AI-disclosure duty on advertisers here. Those two facts meet inside an account your team may not have opened since the last campaign build. This piece gives you the settings path, the check to run on what is sitting on the client's website, and the local obligations that now hang off the answer.

Two Settings That Build Ads From Your Own Website

The two features are marked new in the interface: Landing page images and Sourced videos, which automatically source images and video from an advertiser's existing online content. Search Engine Land's account of the rollout notes that Google can add up to 10 sourced videos to an asset group, against the five-video manual upload limit, and that the settings sit in three categories with a toggle for each feature. Google's own recommended action is to review the Asset optimization section, check whether the two features are enabled, review the content Google may access, and confirm usage rights. (Search Engine Land is owned by Semrush.)

The exposure is wide because the campaign type is. Performance Max is used by over one million advertisers on Google's own count as of April 2025, and agencies are still adding to the format instead of retiring it: campaign volume rose 15.7% year over year in Q1 2026 across more than 21,000 accounts in Optmyzr's State of Google Ads data as reported by Search Engine Journal. Optmyzr sells Google Ads management software, and that account base is its own sample, not the market. Both figures are global, not Philippine.

Machine-assembled creative is also no longer unusual. Most large advertisers now have AI somewhere in the creative process: 83% of advertising executives reported AI deployment in creative in 2026, up from 60% in the 2024 edition of the same study, in IAB and Sonata Insights research covered by PPC Land. That rests on 104 US advertising executives at companies spending at least $1 million a year on media, so read it as direction of travel in a US sample rather than a local benchmark.

Which Toggle Arrives Already Switched On?

In the screenshot that surfaced the change, Sourced videos is switched on while Landing page images and the other visible optimization settings are off, and the advertiser who found it reports that Sourced videos is enabled automatically. Be precise about what that is. One advertiser's report, not a confirmed Google-wide default. The safe reading is the one Search Engine Land itself gives: check your own settings rather than assume the feature is disabled.

A default nobody set behaves like a standing instruction at a bank. It executes on terms someone else wrote, on schedule, until a person cancels it. The difference is that the bank sends a statement.

For an in-house team that is one account to open. For agency teams running multiple client accounts, it is the same unreviewed setting repeated across every account under management, which turns a five-minute check into a portfolio question and a Monday morning job.

Creative sourcing is not the only thing moving inside the console. Google Ads is also rolling out a Translation tool under Tools > Translation, labelled "Translate with Google AI", surfaced on LinkedIn by Jyll Saskin Gales and covered by Search Engine Roundtable. The assembly of ad assets and the translation of them now both happen in the advertiser interface. Whether machine translation holds register in Filipino, Taglish or Bisaya ad copy is a fair question and an open one: we found no published Philippine practitioner evidence either way this cycle, so test it in your own accounts rather than assume a failure.

The ASC Now Puts the Disclosure on the Advertiser

Five weeks before this rollout landed, the local rulebook changed. The Ad Standards Council's Circular 2026-014 on AI in advertising, effective 21 September 2026, requires that ads whose content is AI-created, AI-generated or materially AI-enhanced, "in whole or in part," carry "a clear and reasonably conspicuous disclosure" that AI was used. Accepted labels include "AI-generated," "AI-assisted," "Created with AI" and "Contains AI-generated content," and the disclosure has to be "reasonably noticeable, legible, and appropriate to the format of the advertisement." The scope names advertisers, creative agencies, production companies, media agencies and digital platforms, across text, images, graphics, audio, video, animation and voice. Your agency sits inside that list.

Two lines in the circular matter more than the label itself. The advertiser is "fully responsible for all claims, representations, depictions, endorsements" in the ad, whether made by humans, AI, or both, with "appropriate human oversight and professional judgment" required. And relying on AI "does not excuse or reduce any violation." The ASC can issue a Cease and Desist Order where grounds exist. What the circular does not do is impose pre-screening or re-clearance for AI-enabled ads, and it says nothing about platform asset sourcing specifically. This is self-regulation rather than legislation, and nobody has yet mapped the rule onto this particular platform behaviour.

The supplemental guide dated 2 October 2026 narrows the test, and that test turns on documentation, not technology. "The use of AI in advertising materials does not automatically require the presence of a disclosure," so routine production use (ideation, drafting, editing, compositing, background generation, audio processing) needs none. Disclosure is required when AI is used to "create, replace, or materially alter" ad content that "is covered by, or requires documentation under, an existing law, regulation, permit, consent, authorization, or similar requirement." The worked examples carry the instruction: an AI child in place of a real one who would have needed a DOLE permit, an identifiable person depicted without their actual participation (disclosure does not replace the consent, and the contract or notarised affidavit goes to the ASC for validation), an AI image of a structure that needed the owner's authorisation, and AI music materially altering material that required authorisation. Even the font is specified, Arial or Tahoma at 12 points minimum.

Put that next to Google's own recommended action, the one about confirming usage rights. The Philippine version of that instruction has teeth. Website video and photography routinely carries talent, model, music, stock or property licences scoped to the site rather than to paid distribution, and the obligation sits with the advertiser, not the platform that built the asset.

Identifiable people are the sharpest version of this. The National Privacy Commission holds that generating and sharing AI photos or video from a real person's face or likeness is unlawful absent a lawful basis, because face and likeness are personal information under RA 10173 and generating or sharing such media are "forms of processing of personal data, whatever the nature or purpose of the resulting media," as Philstar.com reported in August. Violations carry one to three years' imprisonment and fines of P500,000 to P2 million. That notice targets synthetic media of real people rather than the auto-assembly of a brand's own licensed footage, so it is not a ruling on Performance Max. Philippine corporate websites are full of staff photos, customer shots and executive portraits, which makes it the question an agency has to be able to answer: does the consent those images were taken under cover paid distribution of a machine-made edit?

Where AI Ad Labelling Actually Bites

Google's interface already answers the labelling question for three jurisdictions. Ads using assets optimized by these features will be labelled as created or edited with AI when shown in the European Union, India and New York State, and text assets on matters of public interest will also be labelled in the EU. Each label maps to a live instrument. EU AI Act Article 50 obligations apply from 2 August 2026, requiring providers to mark synthetic audio, image, video and text so it is machine-detectable and deployers to disclose deepfakes, with Digital Omnibus deferrals pushing marking for systems already on the market to 2 December 2026. India's amendment rules took effect on 20 February 2026, requiring permitted synthetically generated information to carry prominent labels and permanent provenance markers, a duty that binds intermediary platforms, not advertising as such. New York's synthetic performer disclosure law took effect on 9 June 2026, at $1,000 for a first violation and $5,000 for each one after. It is narrower than Google's own labelling: it covers AI-made human performers, and exempts audio-only ads and ads where AI only translates language.

Manila gets no platform label, and that is where the asymmetry now sits. A Philippine audience sees the same auto-assembled asset with nothing on the creative to mark it. The statutory gap is genuine: the proposed Artificial Intelligence Development and Regulation Act was still in committee as of mid-July 2026, with a House technical working group consolidating 26 bills and no advertising-labelling requirement reported in the draft. The self-regulatory gap has closed, though. Disclosure here runs through the advertiser's own judgment about what was AI-made, and through the ASC, rather than through Google's rendering of the ad in a given market. An unreviewed toggle removes exactly that judgment.

Voluntary disclosure stays patchy even among advertisers who use AI openly. Most say they disclose at least sometimes (89%, a marginal increase on 2024) while fewer than half always do, in the same IAB and Sonata Insights study of 104 advertising executives reported by PPC Land. A rule that depends on the advertiser volunteering the information needs a record behind it.

Run the Performance Max Asset Optimization Audit This Week

Open each client's account, go to the campaign settings for every Performance Max campaign, and find the Asset optimization section. Check whether Landing page images and Sourced videos are on. Review what content Google may access. Switch off what the client has not approved. Then run the ASC's documentary test across what is actually sitting on the client's website, which comes down to four questions:

• Identifiable people. Who appears in the site's photography and video, and does their release cover paid distribution of a re-edited version?
• Licensed material. Which assets came from stock, a production house or a talent contract scoped to the website only?
• Property and structures. Do any shots feature buildings, venues or interiors that needed the owner's authorisation?
• Sound and music. Is there audio on the site whose licence stops at the site?

An ownership question is forming behind the compliance one. IPOPHL's draft guidelines on the registrability of AI-related works would protect such a work only to the extent it contains "original expression from human creative contribution," with no fixed test and four categories ranging from AI-assisted to AI-generated output without sufficient human input. The draft is still under consultation and not yet in force, and Director General Teodoro C. Pascua has said it is "not intended to regulate AI generally." Read it as a projection: if the platform assembles the asset, the human contribution could be thin enough that the client paid for creative it cannot comfortably register or defend.

When the platform holds the approval, the thing that still works as an approval trail is a dated, independent record of what ran and what was said about the brand while it ran. A client asking whether they signed off on this gets answered by evidence or by nothing. If the auto-sourced assets are already live, the comparison they will ask for is before and after: what coverage and conversation looked like before platform-assembled creative entered the mix, and what they look like now. That is a measurement question, and a baseline can only be taken once. Read automated sentiment as directional rather than settled, particularly across Filipino and Taglish, and that includes our own scoring. It is the same shift reshaping monitoring practice in the AI era, arriving this time inside an ad account.

Key takeaways

• Check the Asset optimization section of every Performance Max campaign you manage this week. One advertiser reported Sourced videos as automatically enabled, which is a reason to verify rather than a confirmed Google-wide default.
• Treat the toggle as a compliance control. Under ASC Circular 2026-014, effective 21 September 2026, the advertiser carries the AI disclosure duty and full responsibility for every depiction in the ad, and relying on AI does not reduce a violation.
• Work the supplemental guide's documentary test over the client's website before the platform does: people, licences, property, audio. Disclosure is triggered where AI creates, replaces or materially alters content that needed a permit, consent or authorisation.
• Keep a dated independent record of what ran and what was said about the brand during the period. Google labels these assets as AI-made in the EU, India and New York State, so a Philippine audience sees no label and the client cannot tell from the ad itself.

About Media Meter

Media Meter's MediaWatch platform tracks what is running and what is being said about a brand across national, provincial and vernacular Philippine media in real time, which is the dated record to hand a client when the platform made the creative call.