Prompts

"AI visibility tool" and "best AI visibility tool for a small agency managing five clients" are the same topic and two different competitive races. The second names a buyer, a scale, and a use case, and the model will weight its answer accordingly.

What a good prompt looks like#

It is a real sentence. Write what a person types, question mark and all.

It carries the buyer's context. Team size, industry, and constraint change which brands surface. "For enterprise teams" and "for solo founders" rarely return the same list.

It has a stable intent. A prompt asking two things at once produces answers that drift between runs, and your trend line becomes noise.

Group prompts by topic#

Grouping related prompts lets you see the shape of a weakness. If you are strong on brand prompts and invisible on comparison prompts, you have a source problem on third-party pages, not a messaging problem on your own site.

Topic groups also make client reports readable: "we are in the consideration set but lose head-to-head comparisons" lands better than a list of forty individual results.

Keep the set stable#

Rewording a prompt resets its history. Changed wording means a changed question, and the before-and-after are no longer comparable.

Add new prompts freely. Edit existing ones rarely, and when you do, treat the trend as starting over.

How many to track#

Enough to cover the buying journey, few enough to run often. Frequency beats breadth: twenty prompts checked daily reveal movement that two hundred checked monthly will hide. Plans and Credits covers the trade-off in credits.