Find your category's cliche, then break it

Reverse the top ten in your category, read what every one of them shares, and make the one that does not look like the set.

Reversing one winner makes you competitive with it. Reversing all ten and then avoiding what they share is the only version of this that ends up looking like you.

Why this will not look generic: This is the only workflow here that tells you what not to generate, and the list is measured off your actual competitors rather than drawn from taste.

What you walk away with

Ten reversed prompts for the current top ten
The list of choices every one of them shares
One image specified deliberately against that list

Typical run: 10 to 20. Past about twenty the overlap stops changing, so there is no reason to go further.

This one happens in one conversation

You bring: The top ten images for the query you want to win.

Open with: Reverse each of these ten images into a prompt, then tell me what every single one of them shares and which one breaks the pattern.

What the thread holds on to:

All ten prompts at once, which is what makes the comparison possible at all
The intersection it found, so you can keep asking questions of it
The outlier and how it differs, ready to build a brief from

Then you can just ask:

Which of those shared choices is load-bearing and which is just habit?
Write me a prompt that deliberately breaks four of them
Do the same for the top ten in an adjacent category and compare

What the reverse actually gives you

Point it at: The top ten images for the query you want to win, captured one by one.

What comes back (representative example): Ten reversed prompts, read side by side. Every one of them specifies a soft frontal key and a seamless background. Nine place the subject dead centre. Eight land in the same warm-neutral palette band. Seven mention shallow depth of field. That intersection is not the winning formula, it is the category's default, and anything built from it joins the set instead of standing out from it. One image in the ten specifies a hard side key, a coloured ground and off-centre placement. It is also the one you could already picture before you ran this.

Things you would never have thought to type:

ten out of ten on a soft frontal key, as a count
the single palette band eight of them share
the intersection is the cliche, not the formula
which one image breaks the pattern, and exactly how

What you change: Everything in the intersection. Specify against it deliberately, which is a thing you can only do once you can see it.

How it works

Collect the field (In Chrome) - Right-click your way down the top ten results for the query you want to win.
Reverse all ten, then diff them (In Claude) - One thread, ten images. The comparison is the deliverable, not any single prompt.
Make the one that is not like the others - Write the prompt against the overlap: if all ten are soft and centred, specify hard and off-centre on purpose.

Why not just look at them?

Because the thing you need is a count, and counting is exactly what looking is bad at. You can see that ten product shots feel similar. You cannot see that eight of them share a palette band and nine have no tilt.

The overlap is the cliche, not the formula. Copying the winner is the obvious move and it is why categories converge. What every top result shares is the category's default, which means building from it is how you join the set rather than beat it. The useful output here is a list of things to avoid.
You can only diff what has been written down. Ten images are ten impressions. Ten reversed prompts are ten comparable descriptions with the same fields, so the intersection falls out of them. This is the one analysis that is impossible without a reverse engine.
The outlier is the interesting one. There is usually one result that breaks the pattern, and it is usually the one you already remembered. Reversing the set tells you precisely which three or four choices make it different, which is the brief for yours.

Run this before a rebrand, a new category launch, or any time your output is technically fine and still forgettable.

Cost and time

This workflow runs at from 5 credits + 2 reverses + about 1 minute. Your result carries automatically from each step into the next, so there is no re-uploading between tools. Reverses are billed from your reverse meter and generation steps from your AI Credits, both bundled in every pay-as-you-go credit pack.

Browse all guided workflows or all individual tools. Pay as you go, no subscription, AI Credits never expire.