Why the first ideas are always the same
Ask ten people for ideas on a brief and you get roughly the same three territories first. Models do the same thing for a related reason: they are built to produce the most probable next thing, and the most probable idea is the one everybody has already had.
That is not a flaw to prompt around. It is the machine working correctly. The useful question is what you do with the fourth idea onward.
Moves that push past the average
These are the ones that reliably change what comes back.
- Ban the obvious out loud. Name the three territories you already know and rule them out. The space collapses instantly.
- Add a constraint that has nothing to do with the brand. No faces. Must work as audio only. Has to be explicable to a nine-year-old. Arbitrary limits force structural moves.
- Change who is talking. The same brief answered as a rival category, a documentary maker, or a person who hates the product produces different starting points.
- Ask for the version that would get you fired, then walk it back. The territory next to the unusable one is often the interesting one.
- Argue with it. The first response is a position, not an answer. Saying "no, that is a category truth" moves things further than another round of generation.
Knowing when to stop generating
Volume feels like progress and is the most common trap. Twenty ideas is not twice as useful as ten; past a point you are just widening the pile you now have to sort.
The moment to stop is when you have two or three you would defend, not when you run out of appetite. Everything after that is procrastination with a productivity texture.
Then the harder half
Generating is the cheap half. The expensive half is knowing which one survives a client, a jury or a market, and being able to say why.
That is the part AxonCraft is built around: ideas go through a structured critique that scores them and says where they would fail, so what you get back is a shorter list with reasons rather than a longer one. If you want the full run from brief to an approvable plan, see our guide on AI campaign generation.
The reverse problem has its own guide. If you have plenty of ideas but three routes that all say the same thing, see our guide on building creative territories that hold apart.