Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

AI in Marketing › AI Applications in Marketing

Reviewing AI Enhancements in Your Ad Accounts

How to find, preview and decide on the AI enhancements and generated assets in Google Ads and Meta accounts, with an owner, a record and a re-check routine.
Tutorialintermediate9 min readUpdated

What the ad platforms may change or generate in your ads

Platform creative automation comes in two kinds. Assembly combines assets the advertiser supplied: a Responsive Search Ad picks headlines and descriptions for each auction, a responsive display ad combines headlines, images and logos for each placement, and an Asset Group in Performance Max does the same across every inventory type. Generation goes further: the platform alters an asset or creates one the advertiser never wrote. The table summarises what the site's guides record from the help centres. It is a starting point, not a current list.

WhereWhat the platform may doDocumented controlWhere this site explains it
Google Ads, SearchAssemble supplied headlines and descriptions; an optional campaign-level feature creates additional headlines and descriptions from the landing page and existing adsPinning holds a position; the generated assets can be reviewed, removed and the feature switched offResponsive search ads guide
Google Ads, DisplayAssemble supplied assets per placement; enhancements such as auto-generated video and small changes to renderingEnhancements can be switched offResponsive display ads guide
Google Ads, Performance MaxGenerate a video from other assets when none is supplied; generative tools in the campaign builder produce image and text assets from a prompt and the landing pageSupply your own video; generated assets are shown for review before useBuilding an asset group
Meta, Advantage+ creativeStandard enhancements that may adjust brightness and contrast, apply templates, add music or animation, expand images and improve text; generative features that produce background variations, extend images and suggest textReview which enhancements apply to an ad and switch them off individually or as a groupAI for creative, Meta creative testing

Two gaps matter as much as the controls. Neither platform documents how often an enhanced version is shown or how it is chosen, beyond saying the system predicts what will perform; and Meta documents that enhanced versions may be shown to different people, so the ad one colleague sees in the feed is not necessarily the ad a customer sees. Feature names, availability by campaign type and the defaults themselves change, which is why the help centres listed in the sources, not this page, are the reference for what exists today. The steps below say what to decide and record; the help pages say where the switches are.

Before you start

  • Admin or standard access to every Google Ads account and Meta ad account the brand advertises through, including any run by an agency.
  • The brand rules for imagery and voice: how products, people and settings may be shown, and the Brand Voice the copy must hold.
  • The approved claims, ideally a claims library with exact wording and the conditions each claim is approved for.
  • Any wording that must appear in particular ads, such as legal lines or offer terms, and the campaigns it applies to.
  • The AI use register, if the team keeps one, because platform enhancements are AI uses like any other; the AI use register template has a field for features that arrived switched on.
  • A shared sheet for the enhancement register this tutorial builds, kept where the business controls it.

The steps

  1. List every campaign type and ad format where the platforms can change or generate assets

    Start from the accounts, not the features. For each account, list the live and paused campaigns by type and the ad formats inside them: search campaigns with responsive search ads, display campaigns with responsive display ads, Performance Max campaigns and their asset groups, and Meta campaigns whose ads use Advantage+ creative. Include accounts an agency runs for the business. The list defines the scope; an enhancement nobody listed is one nobody will review.

  2. Read each platform's current help pages for the enhancements and their controls

    For each campaign type on the list, read the Google Ads and Meta help pages that describe its enhancements and generative features, and note the date you read them. Write down what each feature does, where it is controlled (account, campaign or ad level) and whether the help page describes it as on by default. Do not rely on last year's notes or on this page's table; the AI in ad platforms guide records that the documentation for generation is still moving.

  3. Record the current setting for each campaign in an enhancement register

    Open each campaign and ad and record what is actually switched on, one row per campaign per enhancement: the account, the campaign, the enhancement, its current state, where it is set and who last changed it if the change history shows it. This is the inventory the rest of the review works from, and it is often the first time anyone has seen the settings in one place. The AI use register template advises checking each tool's settings yourself rather than relying on what people remember switching on, and the same applies here.

  4. Write down what the brand will not let a platform change

    Before looking at any output, write the rules each enhancement will be judged against, so the decision does not depend on whether a particular preview happens to look good. Typical lines: product colour, shape, parts and scale must match the real product; no generated people; required wording must always show; no music on ads designed to be silent; no generated text that adds a claim. The AI for creative guide gives the cases that turn an enhancement into a brand problem however well it performs: a changed product colour, a cropped disclaimer, music added to a silent-by-design ad.

  5. Preview what each enhancement does to real ads

    Use each platform's preview to see what the enhancement produces from the brand's own assets, and look closely at the places generation goes wrong: edges and shadows on expanded images, hands, text inside images, product details and logos. The responsive display ads guide warns that the preview shows a sample of possibilities rather than a fixed design, so look at several renderings and at the smallest and largest formats. The image production tutorial has the review gate for images the team generates itself; apply the same eye to images the platform generates.

  6. Check generated text against the claims library and the landing page

    Read every generated headline, description and text variation as if a new copywriter had written it. Each statement of fact needs Substantiation: either it matches an approved entry in the claims library or it is plainly true of the landing page it points to. Generated text drawn from a landing page can turn "helps reduce" into "eliminates" or lift a feature from one product onto another. Remove what fails, and if the same kind of failure keeps returning, the feature is a candidate for switching off. The AI-assisted ad copy testing guide covers the claim checks for copy the team generates on purpose.

  7. Decide to keep, switch off or limit each enhancement and record why

    Decide each enhancement against the written rules and record the decision, the reason, the person who made it and the date in the register. Most of these choices are two-way doors in the sense of the one-way and two-way doors framework: a setting can be switched back, so an enhancement that breaks no rule can be kept and watched. Where an enhancement can put an unapproved claim or a misrepresented product in front of customers, treat the decision as harder to undo, because the ad has already been seen. Where a whole campaign must carry fixed wording, pinning in responsive search ads holds a line in place, at the cost the responsive search ads guide describes.

  8. Watch the asset reports and labels after launch

    Both platforms document asset-level reporting that shows which assets are used and how they rate. Read it for the assets the platform generated as well as those the team supplied, and remove generated assets that pass the rules but add nothing. Check how Meta labels AI-generated or altered content on the brand's ads, and whether any ad falls under its disclosure requirement for digitally created or altered media in ads about social issues, elections or politics; the disclosure and data privacy guide covers disclosure beyond platform rules. If a generated asset reaches customers with an error, the guide to handling an AI mistake sets out the response.

  9. Set the triggers that send you back to the register

    Name an owner for the register and write down the events that start a fresh look: a new campaign or campaign type, a new account or agency, a platform notice or help page change about creative features, and any complaint about an ad. Add a line to the launch checks the team already uses, such as the Performance Max launch checklist and the creative approval procedure, so that a new campaign is not live until its enhancement rows exist. The NIST AI Risk Management Framework describes this cycle in its own terms: map where AI acts, measure what it does, manage the result and keep someone accountable.

The register does not replace the creative review the team already runs on its own assets; it extends it to the assets the platform makes. The AI in ad platforms guide explains how to measure what has been handed to automation more broadly, and the applications hub places this review among the other AI tasks a marketing team governs.

A leather bag maker reviews the AI settings in its ad accounts (illustrative)

Illustrative example — not a real result

Zephulka Leather reviews the enhancements in its Google Ads and Meta accounts

Zephulka Leather, an invented maker of handmade leather bags that sells through its own site, advertises in 1 Google Ads account and 1 Meta ad account. The head of brand runs the review over 2 working sessions with the paid media manager, who reads the help pages and dates each note.

The inventory finds 14 campaigns and 23 enhancement settings, 17 of them on. The preview and claim check turn up 3 problems: 4 generated search headlines call the bags waterproof, which the care guide forbids; an expanded Meta image adds a second buckle to a bag that has 1; and a generated video in a Performance Max campaign uses an old logo.

EnhancementDecisionReason
Search feature that writes extra headlinesOff in all 6 search campaignsRepeated unapproved care claims
Meta image expansionOff on all adsProduct shape changed
Meta brightness and contrastKept, watchedNo rule broken in 12 previews
Meta text improvementsOff on offer ads, kept on othersOffer terms must stay verbatim
Performance Max generated videoReplaced with a supplied videoOld logo; own footage available

Each row records the person, the date and the help page read. 3 months later the team builds a new Performance Max campaign for a spring sale; the launch checklist line sends it back to the register, where the new campaign's 5 settings are reviewed before it goes live.

Common mistakes

  • Reviewing features instead of campaigns. A setting decided for one campaign does not follow the brand into the next; the register has to list every campaign.
  • Judging enhancements on a single attractive preview. The preview shows a sample; look at several renderings and the formats where things break.
  • Reading generated copy for tone but not for claims. A fluent headline that adds a claim is an unapproved claim.
  • Writing the rules after seeing the output. Rules written first stop a good-looking preview from deciding the question.
  • Copying a list of enhancements from an old note or a blog. The help centres change; read them and date the read.
  • Leaving agency-run accounts out. The brand answers for every ad that carries its name.
  • No trigger for a second look. New campaigns and campaign types bring new defaults, and nobody notices without a checklist line.

Frequently asked questions

Should we simply switch off every AI enhancement in our ad accounts?

Not by reflex. Some enhancements change little that matters to the brand, and switching them off is easy to reverse, so they can be judged on what the preview shows. Others change what a product looks like, what the copy claims or whether a required line appears, and those deserve a firm decision against written brand and claim rules. The point of the review is to make each decision deliberately and record it, rather than leaving whatever the default happened to be.

Do we need to label ads that use platform enhancements as AI-generated?

Follow the platform's own rules at the time you upload, as its help centre documents them. Meta documents labelling for AI-generated or altered content and a disclosure requirement for digitally created or altered media in ads about social issues, elections or politics. Beyond platform rules, the question is whether the ad would mislead someone, for example by showing a person or product that does not exist; the disclosure guide on this site sets out when that arises.

Why recheck settings we have already reviewed?

Because the platforms add, rename and re-default these features, and a new campaign or campaign type may start with settings the old review never saw. A decision recorded against one campaign does not carry over automatically to the next one. Tying the review to the moments when something changes, such as a new campaign, a new campaign type or a help page update, keeps it current without turning it into a standing chore.

Terms used on this page

Frameworks

Sources

  1. Google Ads Help — Google Ads Help
  2. About responsive search ads — Google Ads Help
  3. About responsive display ads — Google Ads Help
  4. About Performance Max campaigns — Google Ads Help
  5. Meta Business Help Centre — Meta

Related