Humanswith.AI — 2 to 1,000+ citations in three months (dogfood)

• Started 2026 with two AI mentions across nine engines.
• Ran the standard Humanswith.AI loop on our own brand for twelve weeks.
• Ended at a thousand-plus citations and 15.4 percent share-of-model.
• Beat every direct competitor by five to ten times.
• Zero paid distribution, zero journalist pitching, zero site rewrites.

Every agency pitching AI search has the same problem. The prospect cannot tell whether the playbook works, or whether the consultant got lucky on one client. So we picked the hardest reference: our own brand.

Humanswith.AI started 2026 with two AI mentions across nine engines. No friendly journalist owed us a favour. No incumbent advantage. No category position. If the loop did not move us, we could not honestly sell it.

The plan was the same loop we now ship to clients. Hermes Visibility measured weekly. The output ranked the three closeable gaps. ContentOS produced the citation-friendly assets to close them. Website Agentic Optimization handled schema and crawler hygiene.

Here is the exact sequence we ran:

1. Step 1 — baseline scan (week 1). Hermes queried each of the nine engines. The result: two mentions total. Zero coverage on six engines. The three top gaps were trusted-publication coverage, comparison-page presence, and dev-community visibility.
2. Step 2 — gap ranking (week 2). ContentOS ranked the gaps by closeability and category leverage. The top gap was third-party trust signals. The second was structured comparison content. The third was technical depth pages.
3. Step 3 — evidence production (weeks 3–4). ContentOS produced ten research articles, four comparison pages, and six citation-friendly explainers. All passed the publish-readiness gate before going live.
4. Step 4 — distribution (weeks 4–6). Outreach landed pieces on three trusted publications. Founder LinkedIn cadence picked up. The first measurable lift showed in week five.
5. Step 5 — compounding (weeks 7–10). Each new citation made the next easier. By week eight ChatGPT and Perplexity were quoting us by name. Gemini and Claude followed in week ten.
6. Step 6 — saturation check (weeks 11–12). By week twelve we hit a thousand-plus total mentions and 15.4 percent share-of-model across the nine engines.

The boring version matters as much as the active version. These are the things we deliberately skipped:

• We did not buy ads. No paid search, no LinkedIn ads, no sponsored placements.
• We did not rewrite our own website twelve times. The site is the product. The lift came from the trust graph around it.
• We did not pitch journalists for coverage. Trusted publications picked up our research articles organically once citation density crossed a threshold.
• We did not chase backlinks. Citations and links followed evidence, not the other way around.

Every measurement is from the live Hermes Visibility dashboard, exported month by month. The cohort is the nine engines we measure for every client: ChatGPT, Perplexity, Gemini, Claude, Copilot, Google AI Overviews, Yandex Neuro, Alice, and DeepSeek. Share-of-model is calculated against the top ten competitors in the AI search optimization category.

The 15.4 percent share-of-model figure is the headline. To put it in context, the next-largest player in the category at month twelve was at roughly 1.8 percent. We were five to ten times ahead of every direct competitor we benchmarked.

We treat the dogfood test as the reference case for three concrete reasons:

1. The hardest target. No incumbent advantage means no shortcut. If the loop worked here, it works anywhere.
2. Full transparency. Every number is independently verifiable. Query the engines directly. Count the mentions. There is no client confidentiality buffer.
3. The method scales across categories. The same loop produced wins for GAC in auto retail (Russia, six weeks) and Birdview PSA in B2B SaaS (North America, eight weeks) right after. Different markets, same method.

If you are deciding whether to engage Humanswith.AI on a tier, the dogfood case is the reference. The loop is real. It ships in weeks rather than quarters. The numbers come from a dashboard you can verify yourself.

We do not promise specific engine outcomes. AI systems are third-party black boxes. What we do is show you the closeable gap in your category on the strategy call, before anyone signs anything.