How a B2B SaaS 6x'd AI-referred trials in seven weeks
Their ChatGPT and Perplexity citations were near-zero, and competitors were the default answer for every comparison query.
AI-referred trials in 7 weeks
share of voice on category queries
weeks to first AI citation
Why this engagement stays anonymous
The client asked us not to disclose the company name, the magnitude of the AI-referred lift was, at the time, a competitive advantage they weren’t ready to telegraph. We’re publishing the playbook because nothing in it is proprietary; everything that worked was a deliberate mechanic, not a stroke of luck.
The starting position
Series-B SaaS, ~$18M ARR, mature SEO program with strong rankings on their head terms. They came to us with a specific complaint: their SDRs were hearing the same line over and over on cold calls.
“We already looked at this category. ChatGPT recommended [competitor], what makes you different?”
A small qualitative audit confirmed it. We ran 47 buyer-style queries across ChatGPT, Claude, and Perplexity. The client appeared in exactly 2. Their main competitor appeared in 41.
The bet
AI platforms cite from a narrow set of high-trust sources: G2, Reddit, the platform’s own product page, a small number of independent reviewers, and the long tail of authoritative-feeling third-party comparisons. Most B2B SaaS optimizes for Google with content the platform won’t cite, comparison pages on the vendor’s own domain score poorly because the platform discounts self-referential claims.
We bet on three changes, in parallel:
Source-side citation work. We mapped every page where the competitor was getting cited, G2 alternatives lists, Reddit threads in three subreddits, two analyst reports, four newsletter comparisons. For each, we built a credible, source-appropriate presence: G2 review campaign, a Reddit-native participation strategy (no astroturf, just a named team member answering questions in good faith), and direct outreach to the analysts and newsletter authors with a 30-minute brief.
On-page citation hygiene. We rewrote the client’s product pages to the structure LLMs actually parse cleanly: clear definitions, direct answers, structured comparisons with named competitors, FAQ blocks with schema. The goal wasn’t to manipulate the platform’s training data, that’s a fool’s errand. The goal was to make the page unambiguous when the platform crawled it.
Reddit narrative monitoring. Two days a week, one of our analysts read every new thread in the three target subreddits. When a question came up that the client’s product solved well, we routed it to the client’s named community lead within four hours. No upvote farming, no sock-puppet accounts, just being early to the conversation with a disclosed, knowledgeable presence.
The 7-week arc
Week 1-2: review-campaign brief shipped, Reddit cadence started. Week 3: first new G2 alternative list updated with the client’s positioning. Week 4: first ChatGPT citation observed in a buyer-style query. Week 5-6: the citation pattern stabilized across Claude and Perplexity. Week 7: AI-referred trials hit 6× the baseline.
“We started this expecting a six-month timeline. What we got was a behavior shift that compounded weekly. The SDR script stopped having to overcome ChatGPT’s recommendation, it started referencing it.”
What’s transferable
The pattern that mattered most: AEO isn’t optimizing for one platform’s algorithm. It’s optimizing for the network of high-trust sources every platform crawls. The cite-graph is the actual product of the work; AI platform behavior is downstream of it.
What we’d do differently
We undercounted the value of the Reddit community work in the initial proposal. It ended up being the second-highest leverage motion behind the G2 campaign, but we scoped it as “supplementary.” On the next engagement in this profile, we scope it as Tier-1 from day one.