Say the answer, then earn it. It feels wrong to everyone trained to build an argument, it is how news has been written for a century, and it is the single highest-yield editing change for AI retrieval.
One is a photograph taken months ago that you cannot edit. The other is a search running right now that you can influence this week. Knowing which one is answering decides what work is worth doing.
Sales and marketing software is the hardest category in B2B to be named in, because the market is crowded, the incumbents are enormous, and the comparison content is owned by other people. Here is what works in it.
Audits, projects and retainers, with the ranges we see in the market and the three things that actually move the number. Plus a way to work out your own ceiling before anyone quotes you.
Perplexity searches first and writes from what it finds. ChatGPT answers from memory unless something tells it to search. That single architectural difference explains almost every behavioural difference between them, including which sites get cited.
Pricing is the question buyers ask engines most often about a specific vendor, and the page that should answer it is usually built in a way machines cannot read. Here is what breaks and how to fix it in an afternoon.
Twenty fixed questions, four engines, two runs each, logged out. It takes an afternoon, costs nothing, and tells you more about your AI visibility than most paid audits. The full set, with how to read what comes back.
Retrieval is a funnel that narrows from thousands of candidates to the three passages a model actually quotes. Four stages, each with a different rule, and only two of them respond to anything you control.
A definition is the most retrievable shape of text a company can write, and almost no SaaS company writes one. Here is what separates a glossary page that gets quoted from a keyword dump with a table of contents.
The right answer depends on which stage is failing, and four of the five likely answers point away from a full AEO retainer. Written by an interested party, with the cases where we are wrong.
