Your buyer asks an AI
before they ask you.

By the time a B2B prospect fills in a form, a shortlist already exists and something helped build it. We find out what the engines say when your buyers ask, fix what is making them say it, and re-run the same questions to show the difference.

Book a call · 15 min →See a real audit →
51%of B2B software buyers now start research in an AI chatbot more often than with Google, up from 29% a year earlierG2, 1,076 buyers, surveyed March 2026
69%chose a different vendor than they originally planned, because of a chatbot recommendationG2, same survey
12%of AI-cited URLs also rank in Google's top 10 for the same promptAhrefs, 15,000 prompts, August 2025. Vendor study
58%lower clickthrough rate for the top organic result where an AI Overview appears — a correlation, not a measured causeAhrefs, 150,000 keywords vs 150,000 controls, February 2026. Vendor study

Each figure names its source and sample size. The two Ahrefs studies are vendor research, labelled as such; the buyer survey is G2's. We walk through what the published evidence does and does not support in what the research actually measures.

01

Which prompts does the buying committee actually run?

A B2B purchase involves several people asking different questions: the practitioner asks which tool does the job, the manager asks about alternatives and pricing, procurement asks about security and compliance. We build the prompt set around all of them, not around keywords.

02

What does the engine cite when it answers them?

Usually not you. Most of what an engine reads about a vendor sits on third-party pages — comparison posts, review sites, forum threads. An agency that only edits your own site is working part of the surface.

03

What changes, and who ships it?

A prioritised fix list with the work named: evidence added to the pages engines quote, rendering fixed so crawlers get full HTML, new pages for the questions you currently lose, and coverage off your domain. We ship them as pull requests or CMS drafts you approve.

04

How do you know it worked?

The same prompts, re-run per engine, before and after, repeatedly. Answers vary run to run, so a single reading proves nothing either way.

Category prompts behave differently when the product is software a buyer can trial in an afternoon. The SaaS version of this page covers that, including a published study of how three engines answer a real software category and how many established brands they never name at all.

Book a call · 15 min →What the audit covers →