Short answer: fill in a strategy sheet before you write the first prompt. AI ads sound like AI because the model was handed a demographic, and the fix is a better input, not a better model: a specific tribe, a named enemy, one contestable belief, the transformation, concrete cultural anchors and a one-line bet.
Any model can give you a hundred ads an hour that are grammatical, on-brand and dead. What no model can invent is who an ad is for, what that person is up against and what they already believe. That is the part a human still has to decide, and it is the part most briefs skip. This is the sheet we fill in first.

Page Contents
Why do AI-generated ads sound generic?
Because the model was handed a demographic. “Adults 25 to 45 who want to learn a language” is true, and no hook can be written from it without guessing. The model guesses the average of everything, and the average of everything is the generic you are hearing.
The fix is not a better model or a better prompt. It is a better input: six lines that name a specific person, a specific enemy and a specific belief.
What goes on a creative strategy sheet?
- Tribe. Not an age bracket. A self-image with a fear and a refusal. In the example, the restarter: three apps, one 400-day streak, still orders in English when the waiter arrives, refuses to be the tourist with the phrasebook, and fears being the person who studied for years and froze.
- Enemy. Not “lack of motivation”. A named failure the tribe would recognise. Here, the streak: it measures opening the app, not speaking the language, and the app profits from the difference.
- Worldview. One contestable sentence, not a tagline. “I am not undisciplined. I got fluent at the app instead of the language.”
- Transformation. An identity shift, not a feature. From protecting a number every night to one unplanned conversation that went fine.
- Cultural anchors. The concrete specifics the tribe actually lives in: the 400-day streak screenshot, the 11pm notification that guilt-trips you into a two-minute lesson, ordering coffee in English out of panic, the cousin who moved abroad and just started talking.
- The bet, in one line. Your streak measures the app, not you. You do not need day 401. You need one conversation.
How do you know a strategy line is specific enough?
The test is whether a stranger could write a hook from it without asking you a follow-up question. “Adults 25 to 45” fails. “The 400-day streak screenshot” writes itself. If a line reads like a demographic, it is not done.
Anchors are worth the most effort, because they are what turns a hook from plausible into recognisable. They come from where the tribe actually talks: app reviews, forums, support text, sales and customer calls, and the comment sections under competing ads. They do not come from a workshop.
What changes when you give the sheet to an AI model?
Hand the weak column to a model and you get the ads everyone is complaining about. Hand it the filled-in column and the hooks, scripts and visual ideas come back anchored to something lived: the screenshot, the notification, the coffee order. The output is still AI-made, and it sounds human because the human decisions were made first.
This is also what keeps volume from becoming noise. Once the sheet exists, every concept can be mapped to a cell of tribe, enemy and anchor, and two concepts that land in the same cell are one bet in two costumes. The full method, including the angle grid and the withhold test, is in our guide to making AI ads that do not look like AI.
Where does a strategy sheet fit in the creative process?
Before everything. The sheet decides the bet; hooks, bodies and overlays only execute it. After it come the moves covered in how to build on a winning ad, the test structure in how to structure creative tests for Meta ads, and the reading of results in how to analyze ad creative performance.
If the ads are not working, check the sheet before the creative: why isn’t my ad creative working walks through the diagnosis.
Frequently asked questions
What is a creative strategy sheet? A short document, filled in before any ad is made, that names the tribe, the enemy, the worldview, the transformation, the cultural anchors and the one-line bet. It decides who the ads are for and what they argue, so that the hooks and scripts can be generated against something specific.
Why do my AI-generated ads feel generic? Usually because the brief described a demographic rather than a person. A model given “adults 25 to 45” produces the average of everything. Giving it a specific tribe, a named enemy and concrete cultural anchors changes the output more than changing the model.
How long does it take to fill in? A first pass takes an hour once you have the raw material, which is reviews, support text and customer calls. The slow part is not writing it; it is finding real anchors. Revisit it whenever a batch of ads teaches you something about who the audience actually is.
Do I need one sheet per app or per audience? One per audience. An app with two distinct tribes needs two sheets, because a single sheet that tries to cover both collapses back into a demographic.



