Eight specific tax prompts, twenty-six weeks: a Manchester accountancy goes from 0% to a stable 38%.
A single-practitioner Manchester accountant audited the twenty prompts an SMB client would ask about UK tax before hiring. Eight were worth the work. Six months later, the citation rate had stabilised — and the practice had stopped chasing referrals.
The audit, week one.
The practice — a one-person accountancy in north Manchester serving mostly contractors and small Limited companies — started at 0% on all twenty seed prompts. The competition was sharper than for most trades: a national accounting-software firm's content marketing team was the most-cited source on twelve of the twenty prompts, and their pages were good. There was no point trying to out-publish them.
The narrowing.
We sat with the practitioner for forty-five minutes and asked which of the twenty prompts he could answer in a way the software firm couldn't. Eight came up: "VAT flat rate scheme worth it 2026", "IR35 outside-IR35 contract template", "limited company dividend tax 2026 threshold", "Making Tax Digital deadlines 2026", and four others specific to the kind of contractor work he handled. The software firm could write generic answers; he could write specific ones with actual case numbers from his own clients.
The pieces.
One long-form piece per prompt, monthly. The first three (April, May, June) covered the most-trafficked prompts. The remaining five landed at one a month through October. Every piece quoted a real HMRC consultation reference, an example client's scenario (anonymised), and a specific 2026-tax-year number — the kind of thing the national firm's content team didn't ship.
The citation curve.
The first prompt cited him on ChatGPT by week three. Gemini took until week eight. By week twenty-six, six of the eight target prompts cited him on at least two of the three engines we audit in production; two cited him on all three. Citation rate stabilised at 38% across the twenty-prompt audit set — a strong result for a solo-practitioner trade against a national-brand content team.
What he stopped doing.
He stopped paying for a Google Ads campaign on the same terms (£600 a month, average customer acquisition cost £180). The inbound from the AI surfaces was free and converting at a comparable rate. The Ads spend went into hiring a part-time assistant.
He stopped going to local-chamber networking events. Most of his new clients in the second quarter of the engagement said they'd come from ChatGPT or Claude. The chamber meetings produced one new client in the same window.
He didn't stop posting on LinkedIn. He noted, though, that the LinkedIn posts had become easier to write because each one was now a riff on a piece he'd already published — and the LinkedIn audience often re-asked the engines for the source.
The lesson.
Eight prompts. Six months. One Friday a week. He's not the most cited accountant on the internet; he is, for the eight specific questions where his work was the best work, the most cited in Greater Manchester.
A note on this story.
The practitioner in this piece is a composite drawn from early Editor-tier engagements in the accountancy cohort. The methodology, the prompt-narrowing exercise, and the per-engine pickup pattern are real; the specific Google Ads and CAC numbers are illustrative of the cohort, not a single firm's books.
- Eli Marsden, Editor.