The best practice for Advantage+ Creative in 2026 comes down to one rule: keep the enhancements that change how your ad is delivered, and turn off the ones that change what your ad says or shows. Visual touch-ups, site links, relevant comments and catalogue formats stay on. Text improvements, overlays, music, animation and every gen AI toggle go off until you have tested them on your own account and seen the variations they actually produce.
This matters more than it did a year ago because the defaults flipped. New Sales, Leads and App Promotion campaigns now launch with the full standard enhancement set switched on, so the question is no longer "should I opt in" but "what do I need to switch off before this goes live". Per Meta's own Advantage+ creative documentation, the system creates multiple variations of each ad and serves a personalised version to each person, with no per-variation breakdown in reporting. You see aggregate numbers. You do not see which version of your ad won, or what it looked like.
There are ten standard enhancements running by default, plus an opt-in "test" set that generates net-new imagery. Here is the verdict on each one.
The Advantage+ Creative verdict table
These verdicts assume you produce your own creative and care what runs under your brand name. If you are a small advertiser running one static with no design resource, Meta's edits may genuinely help you. Everyone else, start here.
| Enhancement | What it does | Keep or kill | Why |
|---|---|---|---|
| Visual touch-ups | Adjusts brightness, contrast, colour temperature and sharpness per placement | Keep | Cosmetic only. The message and composition stay yours |
| Site links | Adds links to other pages of your site beneath the ad | Keep | More clickable area with zero change to the creative |
| Relevant comments | Surfaces a relevant existing comment under the ad | Keep (if moderated) | Free social proof, but only safe with active comment moderation |
| Add catalogue items | Pulls related products from your catalogue into the ad unit | Keep for ecommerce | Extra product surface at no creative cost. Check the feed is clean first |
| Enhance CTA | Swaps or restyles the call to action phrase | Test | Low risk, low upside. Fine to A/B on statics |
| Text improvements | Re-pairs and repositions your primary text, headlines and descriptions | Kill | You lose control of which claim runs where. One agency A/B test measured 26% better ROAS with it off |
| Add overlays | Places auto-generated text or price labels onto the image | Kill | Collides with your designed text and ignores text-safe zones |
| Music | Adds an auto-selected track to Reels and Stories placements | Kill on any ad with a voice track | A random track under a UGC voiceover is a mess. Cue your own music in the edit |
| Image animation / 3D motion | Adds pan, zoom or 3D depth motion to static images | Kill | Warped edges and drifting product shots. If a static needs motion, brief a video |
| Expand image | Extends image edges to fill 9:16 and other placements | Kill | Invented pixels next to your product. Export real 4:5 and 9:16 versions instead |
| Gen AI test set (background generation, image expansion) | Generates net-new backgrounds and imagery, no preview, capped under 5% of impressions | Kill | You cannot review what runs. Product accuracy and claims risk on autopilot |
Key takeaway
Split every toggle into two buckets: delivery mechanics and message changes. Mechanics (touch-ups, site links, comments, catalogue items) can stay on. Anything that rewrites copy or generates imagery gets switched off until it has won a test on your account. And because enhancements now default on, this audit happens on every new campaign, not once.
Which Advantage+ enhancements should you turn off?
The kill list splits into three groups: the text layer, the gen AI visual layer, and music. Each fails for a different reason.
The text layer: improvements and overlays
Text improvements sounds harmless. It is the one I would switch off first. Meta re-pairs your primary texts, headlines and descriptions, and repositions text, based on predicted response. The problem is that your copy variants were written as sets. A curiosity headline paired with a proof-heavy primary text reads fine. That same headline re-paired with a discount line reads like a scam. You approved neither combination.
In regulated and claims-sensitive verticals (skincare, supplements, finance, legal) the stakes are higher than tone. A re-paired variation can put a restricted claim next to the wrong image, and Meta's ad review judges the variation it served, not the ad you built. The rejection still lands on your account. This is the "restricted words" trap practitioners keep flagging: copy that passed review as submitted gets recombined into something that does not.
There is also a plain performance case. Metalla, a performance agency that audited all ten enhancements, ran the same ad with and without text improvements on one account and measured a 26% ROAS improvement with the enhancement off. One account, one test, so treat it as a data point rather than a law. But it shows the direction these edits can push when your copy was already deliberate.
Add overlays fails for a simpler reason: your statics already have designed text. Meta's overlay does not know where your text sits, so you get stacked labels and covered product shots. If you want price flashes, design them.
The gen AI visual layer: expansion, animation, backgrounds
Expand image outpaints new pixels around your asset so it fills 9:16. Next to a lifestyle shot that can pass. Next to a product, it invents environment your brand never approved, and occasionally invents product. Image animation and 3D motion have the same flavour of problem: the motion is generic, edges warp, and a static that was composed to be read in two seconds becomes a slow drifting zoom nobody briefed.
The opt-in test set goes further. As Leapbuzz's teardown of the 2026 enhancement set documents, test enhancements generate whole new backgrounds and extract copy phrases as overlays, with no preview before the ads run, capped at under 5% of delivery, and you must accept all standard enhancements before you can opt in. Read that condition twice. To try Meta's most experimental generation, you hand over your entire creative surface. That trade is bad for any brand with an approval process.
Music
Music adds an auto-selected track on Reels and Stories. On a silent static, plausibly useful. On UGC video, which is most of what converts on Meta right now, it drops a random track underneath a voiceover that was paced and mixed on purpose. Kill it on anything with a voice track and cue music in the edit instead.
Which enhancements are safe to leave on?
Visual touch-ups is the clearest keep. Brightness, contrast and sharpness tweaks per placement are the kind of edit a good editor would make anyway, and Meta classes this level of adjustment as minor, below its AI disclosure threshold. Site links add clickable destinations without touching the creative. Relevant comments is a keep with a condition: it surfaces real comments as social proof, so it is only safe if someone is actually moderating your comment section. Add catalogue items is a keep for ecommerce accounts with a clean feed, since it extends a winning ad with products the viewer can act on.
Enhance CTA sits in the middle. Changing "Shop now" to "Get offer" is not going to damage the brand, and occasionally it wins. Test it on statics; do not lose sleep over it.
One process note on all of the keeps: check the toggles in three places. Ad level controls the individual ad. Account defaults live under Business Settings, then Advertising Settings. Catalogue ads carry their own enhancement settings inside catalogue management. Agencies keep finding enhancements running that were "switched off", because they were switched off at one level and on at another.
Advantage+ and Meta's AI disclosure rules
Meta's 2026 policy updates added a disclosure requirement for AI-generated and significantly AI-modified ad content. The practical split: content modified by Meta's own generative tools gets labelled automatically, so if you leave background generation on, your ads can carry the platform's AI marker without you ticking anything. Content you generate outside Meta (product shots from an image model, a synthetic voiceover, AI UGC) must be declared manually in Ads Manager. Minor edits such as cropping, resizing, brightness and contrast are exempt, which is one more reason visual touch-ups is a safe keep while the generation toggles are not.
If your category is health, beauty, finance or anything else where claims get reviewed, treat the combination of enhancement toggles and disclosure rules as a compliance surface, not a growth hack. The cheapest way to manage it is upstream: ship assets that need nothing fixed.
Enhancements remix your inputs. They cannot create them
Here is the part the settings guides skip. Advantage+ Creative is a remix engine. It recombines the copy, images and video you feed it. Meta Marketing Science reports 12% higher CTR for campaigns using Advantage+ Creative, and Meta's stated rationale for defaulting everything on is that most advertisers run weak, sparse creative that its edits genuinely improve. Both things can be true: the enhancements lift the average advertiser and add nothing to a well-fed account, because a remix engine is only as good as the material it remixes.
Advantage+ can recombine what you feed it. It cannot invent a new angle, a new hook or a new reason to believe. The input layer decides your ceiling.
That means the highest-return "Advantage+ setting" is not in Ads Manager at all. It is the diversity of the creative batch you upload: different hooks, different angles, different formats, different creators. This is the same conclusion we reached in our Andromeda creative strategy guide: Meta's delivery systems now do the matching, and they reward accounts that give them genuinely distinct options to match with. Ten variations of one concept is one input. The volume side of that argument, tier by tier, is in our weekly creative quota breakdown, and the wider discipline of treating each ad as a testable hypothesis is covered in what is performance creative.
At Spark this shapes how we prep every batch before it goes near an ad account. Each creative ships as native 4:5 and 9:16 exports so expand image has nothing to fix. On-image text is held inside a safe zone that clears Meta's UI, so add overlays has nothing to add. Hooks are cut per aspect ratio rather than cropped from a master. The working rule we give clients: if an enhancement exists to patch a gap in your asset, close the gap in the asset and switch the enhancement off. No mystery variations, and the test you read in Ads Manager is the ad you actually shipped. That prep standard is baked into how our batch process works, and it is why our pricing is built around batches of distinct concepts rather than re-crops of one idea.
Advantage+ Creative FAQs
Should you turn off all Advantage+ Creative enhancements?
No. The delivery mechanics toggles (visual touch-ups, site links, relevant comments, add catalogue items) are low risk and sometimes useful. The ones to switch off are those that change your message or generate new imagery: text improvements, add overlays, music, image animation, expand image and the gen AI test set. Audit every new campaign, because enhancements now launch switched on by default.
How do I turn off Advantage+ Creative enhancements in Ads Manager?
At the ad level, open the Advantage+ creative section while building or editing the ad and switch off individual enhancements. Account-wide defaults live under Business Settings, then Advertising Settings. Catalogue ads have a third control inside catalogue settings. Check all three, because a toggle switched off in one place can stay on in another.
Do Advantage+ enhancements trigger Meta's AI disclosure label?
Ads that Meta's own generative tools have modified are labelled automatically, so gen AI enhancements such as background generation carry the platform's AI marker without you doing anything. Minor edits such as brightness, contrast and cropping are exempt. AI content you generate yourself outside Meta must be declared manually in Ads Manager under Meta's 2026 disclosure rules.
Do standard enhancements actually improve performance?
On average across all advertisers, probably. On your account, test it. Meta Marketing Science reports 12% higher CTR for campaigns using Advantage+ Creative, measured on Meta's own infrastructure. Practitioner tests can cut the other way: one agency measured a 26% ROAS improvement after switching text improvements off. Accounts with strong creative production tend to see neutral or negative impact from enhancements.
If you want the toggles handled and the input layer fed properly, that is the work we do all day. Browse the rest of the resources library for the testing and benchmark guides, or tell us about your account and we will scope a batch built to feed the machine.