Overview
A mid-market workflow-automation platform moved from being effectively invisible in AI assistant answers to being recommended alongside its named competitors.
Business context
A B2B SaaS vendor in the workflow-automation category, roughly 80 employees, selling to operations and IT buyers across the UK and Europe. Their sales team increasingly heard prospects say they had 'asked ChatGPT' during evaluation, yet the brand was never surfaced in those answers while three direct rivals consistently were.
Starting metrics
- AI citation rate on decision-stage queries
- near zero
- Share of AI answers naming the brand
- roughly 1 in 10, well behind rivals
- Third-party mentions in comparison content
- sparse and outdated
- Structured, extractable product facts on site
- largely absent
Problems identified
- Classic 'AI does not recommend brand' pattern: assistants had no clean, corroborated source describing what the product does or who it suits.
- No AI citations pattern reinforced by the absence of structured, answer-shaped content on the site.
- Weak entity recognition: the brand name was not reliably associated with its category in the models' underlying data.
- Comparison and 'best tool for X' queries were dominated by third-party listicles that omitted the brand entirely.
Actions taken
-
1
Baseline AI visibility audit
Ran the AI Visibility Framework across a representative set of buyer questions on multiple assistants to establish where, and why, the brand was absent.
-
2
Published answer-first capability pages
Rewrote core product and use-case pages to lead with a direct, extractable answer to 'what is it and who is it for', supported by concise, quotable facts.
-
3
Engineered citable reference content
Created a definitions-and-comparisons hub applying the Citation Engineering Framework so assistants had a corroborated, on-site source to draw from.
-
4
Strengthened entity signals
Aligned naming, category language and structured data across the site and key third-party profiles to reinforce the brand-to-category association.
-
5
Stood up AI answer monitoring
Tracked recommendation share across assistants over the engagement so changes could be attributed rather than assumed.
Results
- AI citation rate on decision-stage queries
- moved from near zero to being cited in a meaningful share of tested answers (illustrative)
- Presence in head-to-head comparison answers
- brand now surfaced alongside its named competitors on most core queries
- Sales-reported 'AI told me about you' moments
- shifted from anecdotal to a recurring theme in discovery calls
- Extractable on-site facts
- materially expanded, giving assistants a reliable source to quote
Timeline: roughly 4 months
Lessons learned
- AI assistants cannot recommend what they cannot cleanly read; extractable, answer-first content is the precondition for visibility.
- Entity clarity across the wider web matters as much as on-site copy when models decide whether a brand belongs in a category.
- Monitoring recommendation share from day one turns AI visibility from a guessing game into a measurable programme.