New
A meaningful share of B2B research now happens inside an assistant that summarises rather than lists. Different mechanics, same objective: be present, be credible, be the one recommended.
The problem
The question "what are the best tools for X" increasingly gets answered by a model producing three names and a sentence about each. If you are not one of the three, you never entered the evaluation. There is no impression to measure and no click to attribute.
What determines inclusion is not the same as what determines a blue link. It is whether the systems have a coherent, well-corroborated understanding of what your company is, and whether the sources they trust say good things about it.
That is buildable, and right now it is far less contested than classic search.
You probably recognise
What you get
A weekly panel of the prompts your buyers actually use, run across ChatGPT, Perplexity, Claude, and Google AI Overviews, with your citation and sentiment tracked over time.
Consistent, machine-readable facts about your company across your site, Wikidata, Crunchbase, review platforms, and industry directories. Models are consensus machines; contradictions cost you.
Definitions, comparisons, benchmarks, and specific numbers structured so they can be lifted and attributed cleanly.
Assistants lean heavily on review sites, listicles, and forum threads. We work the sources that actually get cited for your category rather than only your own domain.
Organization, Product, FAQ, and dataset markup that gives machine readers unambiguous facts.
Deliberate decisions about which AI crawlers can access what, with the trade-offs made explicit rather than left to a default robots.txt.
How we run it
Same order every time, because the steps depend on each other. Skipping ahead is the most common reason these programs underperform.
Run the buyer prompt panel and record where you appear, where competitors appear, and what is said about you.
What the open web says your product does, who it is for, and what it costs. Usually at least one significant inaccuracy is propagating.
Consistent entity data everywhere it is published, starting with the sources that get cited most.
Original numbers and clear definitions. Models cite specifics, not adjectives.
Review profiles, comparison sites, and community threads that feed the models.
Weekly re-runs of the prompt panel. This moves faster than classic SEO in both directions.
Questions
Related reading
How AI assistants build vendor shortlists, why entity consistency matters more than page optimisation, and a measurement approach that works without click data.
May 12, 2026
The structure that works, why the section admitting your competitor is better makes the page convert harder, and the legal and ethical lines worth staying inside.
Jun 9, 2026
Sessions and rankings will not defend a budget in a company with a five-month sales cycle. The metrics that will, how to instrument them, and what to stop reporting.
Jun 23, 2026
No cost, no call required first
Send a URL. Four working days later you get a recorded walkthrough of what is holding the site back and a plan for the next two quarters.
Get my free teardown
