Why Meta counts look-alike ads as one creative.

Meta Andromeda is the retrieval engine of Meta's ads system: the first stage of delivery, which narrows tens of millions of eligible ads to a few thousand before ranking models choose what each person sees. It matters for creative because Meta says its system also groups ads that share key visual and thematic attributes and treats look-alikes as variations of one creative, so ten near-identical ads count for far less than ten. Below: what Meta has said, where Ads Manager measures similarity, and how to build distinct concepts and audit for near-duplicates.

Meta's statements were read on its engineering blog, Meta for Business, the Business Help Center, and earnings call transcripts on November 1, 2026. Inference is labeled as inference.

What is Meta Andromeda?

Andromeda is Meta's machine learning system for ads retrieval. Meta's December 2024 engineering post describes it as the first step of a multi-stage system: it selects a few thousand relevant ads from tens of millions of candidates, and ranking models, which Meta now names as GEM, decide which of those to show. So Andromeda decides which of your ads are considered for a person at all.

Meta built it for volume: the same post counted more than a million advertisers making more than 15 million ads a month with Meta's generative AI tools. Meta's April 2025 guidance drew the practical conclusion: the focus has shifted from niche targeting to creative diversification as the main lever for finding relevant audiences.

What has Meta changed since 2024?

It has reached more surfaces, merged with early-stage ranking, and been rebuilt for scale, with explicit creative guidance alongside. Every figure is Meta's own, published without a baseline.

The October 2026 post carries no new advice for advertisers; the working guidance is still December 2025's.

How does Meta group similar ads?

By what they look and feel like. Meta says its ads system analyzes creatives and groups ads that share key visual and thematic attributes. When several ads look or feel alike, they are seen as "variations of the same creative," and learnings and delivery optimizations are shared at the creative level, not only the ad level (Meta, December 2025). Meta's own conclusion is that iterating on one concept is not enough.

Meta has not published how the grouping works or where alike ends. A Meta spokesperson told Marketing Brew in September 2026 that Meta does not discuss how its ranking models represent individual creative assets.

Our inference: ad count is not concept count, and twelve ads that share an opening shot and a message are closer to one creative than twelve. An agency executive in the same piece says a shared look or opening hook is enough, even with different on-screen text: a practitioner's reading, not Meta's.

What is a Meta Entity ID?

"Entity ID" is practitioner shorthand for one of those groups, not a term on any Meta page we read. Ads Manager and the Marketing API show none, and one agency blog reports a Meta representative calling it internal-only (DataAlly). Articles that name a percentage at which two ads become one Entity ID (40, 60, and 65 percent all circulate) are guessing: Meta has published no threshold.

Where does Ads Manager show creative similarity?

In one estimated label per ad set, next to the fatigue statuses. Meta's Creative diversity help page says the metric compares the images and video thumbnails in an ad set to each other, converts their similarity into a score, and labels the ad set High, Medium, or Low by thresholds Meta does not publish. Meta marks it estimated and in development, and says it shows when an ad set needs more varied images and videos to help prevent fatigue. Practitioners report it arrived under Columns, Customize columns, at the end of August 2026 (MarketingConcurrent).

Read literally, that covers images and video thumbnails, not scripts, voiceover, or copy, so two videos with different stories and the same first frame could look alike to it. That is our reading of the page.

Fatigue is a separate signal. Meta's fatigue help page marks an ad Creative limited when its audience has seen it too often and its cost per result climbs above your past ads, and Creative fatigue at twice as much, counting every recent exposure of that image or video, other campaigns on your Page included. Its first fix is a new ad with a materially different image or video. Richer creative insights are in testing and, per Meta's spokesperson, not widely available. A percentage "similarity score," as some guides describe it, is on no Meta page we read.

What does Meta mean by creative diversity?

Distinct assets for different personas or use cases, not variations of one ad. Meta separates two ideas teams often blur:

• Iteration refines one concept. Meta's example: identical visuals with different text calls to action.
• Diversification makes different concepts. Meta's example: a Reels ad made for Gen Z next to a partnership ad with a creator who connects with new moms.

Iteration can feed diversification but should not replace it, Meta says, and its best practices ask for ads that differ in look, feel, storyline, and message, split by audience motivation or use case. Meta Blueprint says visually distinct creatives drive up to 32% higher performance than similar ones in large-scale Meta tests, without naming the metric (Meta Blueprint, December 2025).

How do you build a creative mix that reads as distinct concepts?

Change what Meta says it groups on, the visual and the theme, before the copy and call to action. Our working rule, not Meta's: a concept is four choices, the persona, the angle (the reason to buy), the format, and the opening visual. Two ads are different concepts when they differ on at least two, one of them the opening visual.

1. Start from motivations. Write the three to five reasons your customers buy, each with its persona. That is the concept list.
2. Give every concept its own first frame. A different setting, person, product shot, or format. The diversity label reads images and thumbnails, so the opening visual is what it sees.
3. Mix formats on purpose. Static, video, carousel, and creator partnership ads. One script cut to four lengths is still one concept.
4. Iterate only inside a concept that earned it. Hooks, captions, and calls to action are iteration; spend them on concepts that passed a verdict, the gate in our creative testing framework.
5. Keep a concept log. Persona, angle, format, first frame, verdict. Without it, a team drifts back to the last winner's look.

In Peachblue, the concept log keeps itself. Every creative is tagged across 31 dimensions, including format, hook style, emotional tone, CTA type, and message angle, and the Intelligence page clusters your creatives into archetypes, so you see how many concepts you have run, not how many files you uploaded. For the next round, the Next Creative Brief turns your winners' shared patterns into concepts to make next, with proven hooks, do and don't rules, reference ads, and a prompt block for an image or video generator. Check it against the archetypes so new work extends the mix instead of cloning the winner.

How do you audit an account for near-duplicate ads?

In two passes: literal duplicates first (one asset in many ads), then conceptual duplicates (different files, same idea), checked against Meta's label and your spend.

1. Export at ad level for the last 30 to 90 days, with spend, results, and each ad's creative. One video in eight ad sets arrives as eight rows.
2. Collapse literal duplicates. Re-uploads, re-crops, and resizes are one creative to a viewer, but the export lists each as its own ad. Group them by eye and add up their spend.
3. Label what remains with the four choices. Creatives that match on all four are one concept, however many files they are.
4. Read the Creative diversity column per ad set. Low on an ad set with many ads usually means one concept many times; Low on three genuinely different concepts deserves a second look, since the label is estimated.
5. Find the concentration. If one concept takes most of the spend, the fix is a new concept, not a fourth cut of the old one.
6. Add before you cut. Meta's fatigue guidance suggests keeping the original running while you add a materially different ad, and its creative insights list combining similar ad sets among the fixes.

In Peachblue, pass one happens at sync. Perceptual fingerprinting recognizes the same creative across ads and platforms, even after re-crops, resizes, and re-compression, and the By Creative view adds up every ad it ran in. Ask Agent Peach, or Claude through the MCP server, for top creatives and the ranking comes back by fingerprint group.

In Peachblue, pass two runs on the tags: filter the Creative Library by hook style, format, or tone to see which combinations repeat, and read the Intelligence archetypes to see how many concepts carry the account.

In Peachblue, concentration and fatigue are already measured: Agent Peach's risk summary reports spend concentration, and the Economics fatigue board flags a creative whose CTR decays against its own best rolling week, with the dollars at stake beside every flag.

Why use Peachblue when Ads Manager has a diversity label?

Because the label says an ad set is too uniform, and Peachblue says which creatives are the same, which concepts carry the spend, and what to make next. Meta's label is free and the only view of how Meta's system sees your ad set, so keep it in your columns, but it stops at one estimated grade per ad set, on Meta only.

Peachblue also tells you whether diversifying works: the composite score ranks a new concept next to the winner it should replace, and the Economics view's hit rate shows whether new concepts produce winners or just volume. Meta and TikTok are on every plan from Starter; Intelligence and the Next Creative Brief start on Pro (see plans).

One boundary, stated plainly: Peachblue's grouping is its own. It does not read Meta's internal grouping, an Entity ID, or the Creative diversity label, and it does not create, edit, or pause ads. It analyzes your own ads, not competitors'. More on creative fatigue, on running this audit in Claude for Meta ads, and on what creative intelligence is.