Four steps for
ongoing search work.

Ongoing work starts by agreeing the buyer questions, platform coverage and usable evidence. Brand mentions, owned-domain citations and unavailable captures are kept separate. Search Console and GA4 are reviewed where access or exports exist, alongside crawl access, page relevance and internal links. The next stage repairs confirmed defects. Structured data must fit visible facts; llms.txt is optional documentation, not a Google visibility requirement. From month two the agreed delivery programme runs: four priority page or evidence improvements at Citation Engine, eight at AEO Stack, or an agreed Search Lab backlog. Thomas Howcroft reviews the work. On the 1st of every month a signed PDF report shows the citation index, organic movement, what shipped, and what is planned next. The tooling speeds the work up. The judgement calls stay human.

First agree the buyer questions, named platforms, answer modes and capture method covered by the retainer. Thomas checks whether a reader can immediately tell what each priority page is about and what service it offers, then compares that answer with buyer intent and relevant search data. A provider is checked only where the agreed route and usable access exist; unavailable observations stay visible. An official API answer is not interchangeable with the consumer assistant. Google AI Overviews is excluded from this API workflow. Review usable Search Console and GA4 evidence alongside the live site, then separate confirmed defects from explanations that still need testing.

What ships

• Dated answer baseline with platform coverage and unavailable rows
• Technical audit of access, indexing and relevant page facts
• Mentions and owned-domain source citations recorded separately
• Priority list ranked by evidence, impact, and effort

Prioritise crawl access, indexing evidence, readable content, useful internal links and accurate business facts. Give each priority page one main question to answer quickly, then cover closely related questions in depth. Expand an existing page when the topic belongs there; create another page only for a distinct buyer need. Check headings and performance without treating a Lighthouse score as a citation predictor. Add or correct structured data where it fits visible content and a supported feature; Google requires no special AI schema. If maintained, llms.txt is optional documentation: Google ignores it for Search visibility and rankings, and other services may ignore it too. Verify each repaired condition without presenting it as a guaranteed citation gain.

What ships

• Priority access and indexing findings resolved or handed over
• Relevant structured data checked against visible content
• Optional llms.txt checked for accurate public links
• Sitemap and internal links checked against intended pages
• Repair evidence and unresolved limits recorded

Citation Engine includes four priority page or evidence improvements a month; AEO Stack includes eight. Search Lab uses an agreed delivery backlog. Work may improve an existing page, create a page for a distinct buyer need or strengthen the evidence behind a useful claim. Length and format follow that need. Each item starts from available search data, answer observations and relevant competitor coverage. Claims need appropriate public evidence or permissioned examples. Thomas Howcroft reviews the content, internal links, relevant structured data and usefulness before publication is authorised. Agreed search and answer checks then record what changes, without attributing every movement to the edit.

What ships

• 4 priority improvements at Citation Engine; 8 at AEO Stack; an agreed Search Lab backlog
• Useful page or evidence work checked against the brand voice and visible facts
• Relevant internal links updated with delivery
• Dated follow-up observations at the agreed checking cadence

On the first of every month a signed PDF report lands in your inbox. It covers the citation index, organic movement, what shipped, and what is next. Relevant buyer questions can be checked weekly during an active improvement run, with the named platform, date, answer, brand mention and cited URL kept together. Repeated observations are useful, but they do not guarantee a citation or prove that one page edit caused it. Thomas reviews the evidence, adjusts the priorities, and locks the next month. The tooling speeds the work up, but the calls are still human.

What ships

• Signed monthly PDF report on the 1st of every month
• Citation index updated when official-API checks run
• Next month's plan locked in writing
• Quarterly or monthly strategy review (depending on tier)

• Async-led. No bloated discovery process. No standing meetings. Written reports and screen recordings do most of the work. If a call sharpens the decision or the strategy, you speak directly with Thomas.
• Engineer-led. No account managers. No project coordinators. You email Thomas Howcroft. He replies. Nothing passes through three layers.
• AI-assisted, human-reviewed. Six AI-assisted workflows under one human operator. Atlas tracks. Forge drafts. Compass plans. Scribe reports. Hawk watches competitors. Anvil keeps the technical surface healthy.
• Transparent. Every shipped piece, every citation gained or lost, every technical change is in the monthly report. No black boxes. No mystery deliverables.
• Capped roster. A premium roster means the operator has time to think about each client. No portfolio dilution.
• Cancellable. Citation Engine and AEO Stack have a 6-month minimum; Search Lab Retainer has a 12-month minimum. After the applicable minimum, cancel with 30 days written notice. The published package terms include the citation share commitment and audit credit conditions.

Search Lab is TomHow's method for measuring search visibility, AI citation evidence, page gaps, source quality, and next-step priorities. It is not a separate software product.

Search Lab starts with the customers a business wants to reach, a dated baseline and a focused set of search and AI-answer questions. Findings become a priority list of page, technical and evidence improvements within the agreed scope. Delivery records show what was completed; follow-up observations show what changed. Confirmed enquiries and qualified opportunities need separate evidence. The public sample report demonstrates the format with fabricated data. A repeat check does not prove that one intervention caused a result.

TomHow separates verified evidence, anonymised examples, fabricated sample data, directional patterns, and hypotheses. That keeps public pages useful without exposing private client data.

• Sample reports are clearly labelled
• Client proof needs approval before publication
• Private account exports stay private
• Claim strength follows the evidence
• Reports explain what the data cannot prove

Client data stays private. Public samples are fabricated or anonymised unless a separate client-approved case study exists.

Every retainer client receives a signed monthly PDF on the 1st. Same format, same source discipline, same level of detail.

• Executive summary with month-on-month KPIs
• Visibility snapshot across AI search and SEO
• Dated answer observations for agreed providers, with mentions and source citations separate
• Movement, blockers, next actions, and shipped work
• Signed note and source index

The report is built for decision-makers, not SEO dashboards. It explains where the brand was cited, which searches improved, what shipped that month, what is blocked, and what will happen next. One document, signed, archived, and easy to forward internally.

Platform coverage follows the agreed retainer scope. API answer checks record mentions and source citations separately and do not reproduce every consumer assistant answer. Available Google search and AI impression data is reported separately. Stored enquiries, delivery status and qualified opportunities need their own evidence; a contact click is not an enquiry.

The public example is deliberately fabricated. It uses a sample company, sample prompts, and sample numbers to show the format without exposing any real internal, affiliate, or client report data. Real reports stay private unless a separate client-approved version is created.

Fabricated sample data demonstrates the report format. It is not a client result. Open the sample.

Rule-based builders for llms.txt and JSON-LD schema. Open the builder or open the schema generator.

GBP 2,250. A signed diagnostic with linked evidence rows, ten prioritised fixes and a 90-day action plan. Credits back in full against a qualifying retainer signed within 30 days. See the audit terms.