Cybersecurity SaaS (EDR)

A Series B EDR platform selling into mid-market IT and security teams. The product had substance, but the brand was missing from the threads where shortlists get built: comparison posts, incident post-mortems, migration regret threads, and 'what should we buy' questions.

When we started, brand presence sat at 4 of 68 tracked Reddit URLs (6%). Across the first 14-day reporting window, zero AI chat answers in our tracked prompt set named the brand. Competitors were named in 73% of the same answers.

We treated this as a credibility problem, not a distribution problem. We started by mapping the recurring questions that precede EDR evaluation, then built a posting and commenting program around the threads where those questions land.

Every post and in-thread response was designed to hold up if you removed the brand name: incident response lessons, deployment failure modes, real tradeoffs between architectural approaches, and field notes from production environments. The work was client-approved before publishing, with strict guardrails (no fake accounts, no vote manipulation, no shortcuts).

In comparison threads, we participated without hijacking. We answered the buyer's actual question, named tradeoffs honestly, and stayed away from the 'we do everything well' posture that triggers skepticism in technical audiences.

Paid amplification was used only on angles that survived organic scrutiny first, pointing to evaluation assets (benchmarks, deployment guides, cite-able documentation), not to hype landing pages.

Engagement ran for 150 days, structured as 10 rolling 14-day reporting cycles. Each cycle produced a standardized report covering the five metrics above plus the top Reddit URLs by AI retrievals and a sample of AI answers where the brand was named.

Brand presence is defined as the number of tracked Reddit URLs (the URLs AI chats pulled as sources in the period) where the brand was either mentioned in the AI's answer text alongside that URL, or where we had placed a comment at any point. The numerator and denominator are both observable, not modeled.

AI answers naming the brand is counted from a third-party AI search monitoring layer that runs a fixed prompt set against ChatGPT, Gemini, and Google AI Overviews on a daily schedule. The count is exact, not estimated, and the breakdown of how many of those answers also cited a Reddit URL is taken from the same source.

Lead attribution uses a self-reported 'Where did you hear about us?' field on the post-sale form, with 'AI / LLM (ChatGPT, Claude, Perplexity)' and 'Reddit' as explicit selectable options alongside the usual choices. Close rate is reported only for leads who selected one of those two.

What this generalizes to: any B2B category where the buyer asks the internet before they fill in a demo form. If your prospects pressure-test tools in threads, the program described above moves both AI presence and pipeline. The five-metric reporting structure makes that movement visible week by week, not in a 12-month retrospective.

What this does not prove: that posting volume alone produces the result. The 23 actions per cycle figure is meaningless without the quality bar (client approval, defensible specifics, willingness to name tradeoffs). Programs that scale activity without scaling that quality bar will not see the same AI answer movement, because LLMs preferentially cite content that holds up to scrutiny.