How to report SEO without lying to anyone

SEO reporting has a credibility problem, and it is self-inflicted. Reports lead with numbers that move on their own, attribute everything to the work, and never mention the things that did not happen.

A more honest structure is not harder to produce. It is just less comfortable.

Lead with what changed and why

One paragraph. What moved, what caused it, what you are doing next. If you cannot write it, the analysis is incomplete and the charts will not rescue it.

Separate what you did from what happened

Two sections, deliberately not merged:

Work completed — the actual changes, with dates.

Outcomes — the numbers, with the same date range.

Keeping them apart makes the connection arguable, which is correct, because it is arguable. Merging them implies causation you cannot demonstrate.

Report the things that did not work

A page you optimised that did not improve. A prediction that was wrong. This is the single strongest credibility signal available, and it costs nothing except discomfort.

Nobody believes a report where everything worked, and reasonably so.

Use metrics that cannot flatter you accidentally

Impressions for visibility, segmented.

Clicks and click-through rate for the specific queries you worked on, not sitewide.

Queries crossing from page two to page one — a direct count of the thing you were trying to do.

Conversions or leads where measurable, with an honest note about attribution limits.

Avoid: sitewide average position, total keywords ranking, domain authority scores from third-party tools, and anything that improves when you lose visibility.

Name what you cannot see

Consent-driven data gaps. Attribution windows. Search Console's sixteen-month retention. Queries Google withholds for privacy.

Stating limitations makes the rest of the report more believable, not less. A report with no acknowledged limits is a report nobody has checked.

The carry-forward

Last month's recommendations, with their status: done, not done, done and it worked, done and it did not.

That single section converts reporting from a monthly artefact into a process, and it is the only mechanism by which anyone ever finds out whether the recommendations were good.