Stabilized a stalled account, then rode an LLM-traffic surge most of the category hasn't caught up to.

I picked up this account early in the engagement after the prior strategist exited. The retainer was at risk: the client had no reason yet to trust a new name on the account, and the site itself was a different kind of problem: a programmatic page set in the hundreds, built up tool by tool and blog by blog over years, with no standing methodology for deciding what to refresh, what to consolidate, and what to kill.

That's a trust problem and a scale problem at the same time. You can't propose a credible plan for 929 pages by eyeballing them, and you can't rebuild trust on a stalled account by asking for more time before showing any read at all.

Stabilizing the relationship and auditing the site weren't separate workstreams: the audit was how I rebuilt the trust. Showing up with a structured, page-by-page read across the entire tool and blog library, fast, was the proof of competence the account needed.

I self-initiated and built the audit methodology with Claude: a repeatable pass across all 929 pages, grouped into 45 topic hubs, scoring each page on overlap, search performance, and structural gaps rather than reviewing them one at a time. The output wasn't a one-off report: it's a methodology I now reuse across other accounts.

From there: refresh the pages with real search demand and thin content, consolidate the overlapping ones, and keep a standing cadence of new tool and blog pages aligned to what the audit showed was missing. The retainer expanded into a dedicated content package once the first read landed.

Two months ago, LLM-referred traffic on this account started compounding, and the audit wasn't the reason. A separate workstream ran in parallel: stripping stale 2025 date callouts out of the blog, rolling schema out sitewide, rebuilding header structure across every tools page, and shipping 20+ pieces of new and refreshed content. The site was already showing up in LLM answers before any of that. What changed was ChatGPT 5.5 adding more inline citations to its responses that same window, turning visibility that already existed into traffic that could actually be attributed. April: 11,639 sessions from LLM surfaces. May: 52,786, up 353.5% in a single month. By mid-June it was already at 80,363 month-to-date, on pace for roughly 134,000, another 153.8% on top of May.

Organic subscriptions told a related story: May closed at 2,548, up 38% month over month and the highest monthly total since October 2024. June was already pacing to beat it again by the third week of the month.