Prompt space
Prompt space is every way people could ask an AI platform about your category. You track a sample of it, so the sample needs to be fair.
Prompt space is every way a question about your category could be asked. Each subject, each kind of buyer, each wording, each place and language adds more. For most brands it runs to thousands of questions, far more than anyone could track.
So you track a sample. The numbers you read are only as true as that sample is fair. If you pick the questions you already do well on, you'll feel good and learn little.
How big it gets
Take one need, like a coffee subscription. People ask about price, about roast, about gifts, about decaf, about delivery to their town. They ask as beginners and as experts. They ask "best," "cheapest," "is it worth it" and "X or Y." Multiply that by every market you sell in, and the space grows fast.
Picking a fair sample
Aim for a sample that looks like the real space, in miniature:
- Spread your prompts across the subjects your customers care about
- Mix questions from people learning, comparing and buying
- Keep most of them free of your brand name
- Weight them towards the questions people ask most
For example, Harrow & Pine could track 30 prompts on loose-leaf tea, all about premium gift sets, where it's strong. Its visibility would look high. If most people in its market ask about everyday tea, and it rarely comes up there, those 30 prompts hide the real picture. Moving 15 of them to everyday questions gives a lower number and a truer one.
Keeping it fair over time
The space changes as people find new things to ask. Check your sample now and then against what people ask, and replace prompts nobody asks any more.
In AnswerOrbit
Your plan sets how many prompts you track, so the sample has a size limit. Generate prompts in Topics and prompts lets you choose the mix of learning, comparing and buying questions. The Demand tab shows which questions people ask most, and the Coverage tab lists questions you don't track yet.