Editorial & corrections policy.

Source discipline

Every recommendation in an audit carries a source. We use a four-grade system consistent with the master handbook this business is built to:

• Grade A — official Google, Bing or platform documentation. Stated as fact.
• Grade B — strong independent evidence (peer-reviewed research, large original datasets, established engineering docs). Stated with the study or source named.
• Grade C — reasonable inference from A or B evidence. Stated as "likely" or "consistent with".
• Grade D — experimental. Only used with a test design attached, and never on money pages of live customer sites.

Where a recommendation depends on judgement about the specific business (its market, competitors, or goals), we say so and cite the customer's own stated intake as the reference.

What we will never do

• Invent an issue to reach the three-ticket count. If only two fixes pass our checks, we deliver two and refund in full — see the refund policy.
• Recommend link buying, guest-post schemes, "domain authority packages", PBNs, or reciprocal-link deals. These are Google spam-policy violations and cause harm.
• Recommend rewriting a page that is already ranking for a query, on the theory that "AI needs different content". Controlled research shows citation-style body rewrites can reduce retrieval — we protect what is working.
• Publish a customer's report, quotes, numbers or private data without written consent. Reports live on noindexed, access-controlled URLs.
• Present a lab performance score as if it were field data — a Lighthouse number is not a CrUX number, and we say so on every ticket.
• Fabricate credentials, testimonials, reviewer names or results. We do not present AI output as the work of a human expert.

How AI is used, and how it isn't

The audit is generated by AI (Anthropic's Claude) and delivered without human review. We disclose that on the site and on every report. Automated rules decide what the AI may and may not do:

• The AI does: choose the three highest-impact issues from the crawler's findings; write each fix with exact recommended wording; explain why it matters, with a source grade.
• Software blocks any fix that: names a page not found in the crawl; quotes "current" text that isn't on the page; asserts facts about your business you didn't give us; promises rankings, traffic or AI citations; or recommends a tactic against Google's spam policies.
• Google's own scaled-content spam policy applies regardless of authorship. We follow it.

The correction process

1. Write within seven business days of delivery to the email address on the report. Include the ticket number and the fact you believe is wrong, with evidence where you have it.
2. We respond within one business day acknowledging the correction request and confirming whether it will be treated as a factual correction (rewrite the ticket, no cost) or a change of opinion (out of scope for the correction round).
3. If we agree it's a factual correction, the corrected ticket replaces the original on your report page within three business days. We keep both versions in our records for our own reference.
4. If we disagree, we say why in writing. If you're still unhappy, we refund in full to close the matter — we would rather refund than leave a customer feeling ignored.

Owner-side content: locked terms and the URL-preservation gate

The master handbook this business is built to has strict rules about preserving the titles, H1s and URLs of pages that are already earning impressions. We apply the same rules to your site: no ticket will recommend rewriting a title, H1 or URL of a page currently ranking without documenting the before-state, the mechanism for the change, the expected effect size, and a roll-back plan. See the sample audit — every ticket includes the roll-back plan.

Complaints and escalation

If a complaint cannot be resolved by correspondence, the owner named on about is personally responsible for resolution. Statutory rights under the UK Consumer Rights Act 2015 apply and are not limited by anything on this site.

Reviews of this page

This editorial policy is reviewed quarterly (or sooner if Google's own documentation on any of the topics above changes). Substantive changes carry a dated note at the bottom. This page was last reviewed on 23 September 2026.