The 5 patterns that give away generic-AI ads
AI made producing ads free. It also made a certain kind of ad instantly recognizable — and the Meta auction quietly taxes every one of them. If you run ads for a DTC brand, you’ve seen these five patterns in your own account, probably this week. You may have shipped a few. We have too.
None of them are about which model wrote the words. They’re about what the writer didn’t know: what stage your market is at, which hooks your audience has already scrolled past a hundred times, what your customers actually say when they complain. A tool that only writes will happily produce a hundred grammatical variations of the same wrong idea. Here are the five tells, the mechanism behind each one, and the proven move that replaces it.
The interchangeable opener
“Say goodbye to frizzy hair. Say hello to effortless confidence.”
How to spot it
Run the swap test: replace your product with your closest competitor's and read the ad again. If nothing breaks, you don't have an ad — you have a category placeholder. The tells are openers built from stock verbs: “Say goodbye to X”, “Unlock”, “Elevate”, “Discover the secret to effortless Y”.
Why it costs you
The hook decides in the first one to three seconds whether anyone stays. An opener that could belong to any brand carries no pattern interrupt and no specific claim to process, so the thumb keeps moving. Low early hold time tells Meta the creative is weak, and the auction responds the only way it can: your CPMs creep up and your reach quality drops. You end up paying more to be scrolled past faster.
What winning brands do instead
Winning brands don't open with vibes — they open from a named formula and adapt it. Pain Mirror: “Why your [routine] isn't working (and it's not because of [the obvious reason])”. Reverse Logic: “Why washing your hair less made it greasier”. Or the hundred-year-old classic that still prints: “Do you make these mistakes in [area]?”. These formula families — curiosity, pain, proof, promise, identity — are old direct-response scaffolding, and they work because each one forces a specific claim into the first line. Test three to four hook variants per concept, not one.
“Why your “hydrating” shampoo is the reason your hair is frizzy by 3pm.”
Benefit soup
“Effortless. Sustainable. Premium. Designed for your everyday.”
How to spot it
Three or more benefits per breath, adjective stacks where a claim should be, and not a single number anywhere. AI tools produce this because listing every brief input feels “complete” — but completeness is a spec-sheet virtue, not an ad virtue.
Why it costs you
Vague claims slide off the brain; there is nothing to grab. When everything is a benefit, nothing is the benefit, and the reader leaves holding zero of your five points. Specificity is also a truth signal: “22.7%” reads more credible than “over 20%”, and “in 41 seconds” beats “instantly” — vagueness doesn't just dilute, it quietly reads as lying.
What winning brands do instead
Concentrate. Run each feature through feature → benefit → benefit-of-the-benefit and pick the single sharpest cut, then make it concrete. Take a balance board: “200+ tricks to master” is the feature list, and “kids go absolutely wild on it” is one emotional cut. Both describe the same object, but only the second one is a reason to buy. One sharp edge, oddly specific, tends to outrun five polished adjectives — and unlike an adjective stack, a single claim is something a test can actually settle.
“One t-shirt, 143 wears so far, still no pilling. Here's the boring reason why.”
The closing message on a cold audience
“20% OFF this week only — shop the collection now.”
How to spot it
An offer creative — discount, urgency, “shop now” — running in prospecting, aimed at people who don't yet know they have the problem your product solves. AI tools default to offer-speak because the brief says “promote the sale”, and a sale is all the context they have.
Why it costs you
Eugene Schwartz mapped this sixty years ago: an audience sits at one of five awareness stages, from unaware of the problem to ready-to-buy, and each stage responds to a different message. A discount is a Most Aware message — it converts the small slice who already decided and bounces off everyone else. Fired at cold traffic it produces the signature every media buyer knows: retargeting looks great, prospecting bleeds, CPMs high, CTR low. The ad isn't bad. It's at the wrong stage.
What winning brands do instead
Match the message to the stage, and scale from the most-aware ring outward. Cold audiences get problem-surfacing and education — longer copy, native-feeling creative that names what they're feeling before naming your product. Solution-aware traffic gets the mechanism. The discount static goes only to the bottom of the funnel, where it belongs. Same product, five genuinely different ads — which is the diversification that actually matters, not five fonts on the same offer.
“Often tired by mid-afternoon and blaming your sleep? It might be the low-grade dehydration you've stopped noticing.”
The claim with no mechanism
“The most advanced hydration formula ever created.”
How to spot it
A superlative claim in a category where every competitor makes the same one. “Most advanced”, “clinically proven”, “next generation” — claims that were maxed out years before you entered the market.
Why it costs you
Markets mature. Once your audience has heard every claim amplified to its ceiling, a bigger adjective isn't persuasion — it's noise, and it actively breeds skepticism. A promise alone has been heard thousands of times. A promise plus an unfamiliar mechanism is a different object entirely: it gives the skeptical brain a how to hold on to, which is where credibility and curiosity come from in a sophisticated market.
What winning brands do instead
Name the mechanism, and aim it at the root cause. Not “hydrates deeper” but “standard hyaluronic acid sits on top of your skin — the molecule is too large to absorb. Ours is split into three molecular weights, so it works at three depths.” That is the stage-three move in Schwartz's sophistication model, and it's what winning brands in crowded categories run almost exclusively. And because mechanisms get copied, they keep challenging their own: when every competitor says “three weights”, the mechanism gets rebuilt, not the adjectives.
“Your moisturizer isn't failing you. Its molecule is just too big to get where the dryness starts.”
Product language instead of customer language
“Our innovative design combines functionality with style for the modern consumer.”
How to spot it
The copy reads like the spec sheet wearing a smile. “The modern consumer”, “elevate your routine”, “crafted with care” — sentences no human has ever said out loud about a product at 11pm. Grammatical, polished, and completely interchangeable. This is the actual “AI flavor” people can smell but can't name.
Why it costs you
It happens because the copy was generated from product attributes instead of mined from the market. The strongest ads work through recognition: the reader sees their own half-formed thought written down and feels found. Copy derived from a spec sheet can't be recognized — so it doesn't get felt, saved, screenshotted, or clicked. It also can't meet customers where they've already been burned, which is where most buying decisions actually live.
What winning brands do instead
Mine the verbatims. Reviews, support tickets, DMs, Reddit threads — the exact sentences customers write, in their words, spelling quirks included. Build the hook out of one: “I've bought four of these and they all pilled after two washes” is a better first line than anything a model invents from your brief. Then run the failed-alternative move: “You tried X. It didn't stick. Here's why it couldn't have worked.” Empathy for the past failure, then the mechanism. An ad built on a real verbatim can never fail the swap test — it's already wearing your customer's voice.
““I've bought four white tees this year and every one pilled by wash three.” — the review that made us redesign the knit.”