Four real files, not a page describing one. A seven section monthly report, a findings tracker with a decision and an owner on every row, and a KPI sheet that states where each number came from.
An SEO report should be seven sections and three pages, not thirty. What changed, the numbers behind it, what was fixed, what was published, what is open and who has it, what you are deliberately not doing, and what happens next. The sixth is the one no template has and every client relationship needs.
No form, no email, no account. Edit them, rebrand them, send them to a client as your own. Everything on this page is also rendered below, so you can read the whole template without downloading anything.
Seven headings, each with the prompts that sit under it and the thing that does not belong in it. The Word and Markdown files are exactly this, with the guidance as italic lines you delete as you fill it in.
Three to five sentences a busy reader can stop after. If they read nothing else, this is what they know.
Leave out. A recap of what SEO is, a list of everything that was touched, and any sentence that would be true in any month of any engagement.
The small set of figures worth tracking, each with the date and the tool it came from. Use the KPI tab of the spreadsheet.
Leave out. Rankings for keywords nobody searches, a domain authority score, and any metric that has never once changed a decision in this engagement.
Shipped work only. One line each, linked to the finding it closes in the tracker.
Leave out. Work in progress, work planned, and anything where the fix is written but not deployed. Those belong in section 5.
New and updated pages, with the reason each one exists. A page with no reason is the thing to catch here.
Leave out. Word counts, publishing cadence as an achievement, and a list of pages with no statement of what any of them is for.
Everything found and not yet fixed, with a name against it. This is the section that makes the report a working document rather than a summary.
Leave out. The full findings list. Link the tracker instead; a report that reprints 200 rows is the report nobody opens.
The section no template has. Everything deliberately deprioritised, with the reason, so nobody has to reconstruct the decision in six months.
Leave out. Silence. A thing left undone with no reason recorded reads as a thing forgotten, and it will be raised as one.
Three to five items, in order, with the largest first. Anything longer than that is a list rather than a plan.
Leave out. A rolling backlog copied forward each month. If an item has appeared three times without moving, it belongs in section 6.
The third column is the one other templates leave out, and it is the one that stops a figure being over-read in a meeting. A simulated lab score reported as what visitors experienced is the most common version of that.
| Metric | Where it comes from | What it does not prove |
|---|---|---|
| Clicks from organic search | Search Console, Performance report, Web search type | Reported on a delay of a couple of days and deduplicated by property. It is not sessions and will not match analytics. |
| Impressions | Search Console, same report | An impression is not a reader. Rising impressions with flat clicks usually means being found for things you do not deserve to win. |
| Queries with impressions and no clicks | Search Console export, filtered | The most useful row here. It does not tell you why, only where to look. |
| Average position | Search Console | An average across every impression, so it moves when your query mix changes and nothing about your rankings did. |
| Indexed pages | Search Console, Pages report | A count, not a health measure. A site can index everything it has and deserve none of it. |
| Enquiries or conversions from organic | Your analytics, with the attribution model named in the cell | Attribution is a model and every model is wrong somewhere. Name which one, and compare like with like over time. |
| Core Web Vitals, field data | Chrome UX Report, 75th percentile, 28 day window | Only exists once a URL has enough real traffic. If there is no field row, do not substitute the lab number. |
| Core Web Vitals, lab score | Lighthouse or PageSpeed Insights | A simulation on a shared machine. It moves between runs on an unchanged page, so never report a single reading. |
| Referring domains | Whichever backlink index you use, named in the cell | Every index sees a different web. The trend within one tool is meaningful; the absolute number is not comparable to anybody else's. |
| Assistant mention rate | Your own prompt set, run against each assistant, fresh conversation per question | A rate over a prompt set, not a ranking. Re-run it: these systems are not deterministic and one run is an anecdote. |
Twelve columns, and every fixed-value one is a dropdown in the workbook. There is deliberately no “noted” status: a row with nobody’s name on it is a row nobody will action, and recording it anyway is how a report gets delivered and never opened.
The files open with these already in them, sorted by how often the thing is actually broken rather than by category. Importance is an opinion and a rate is not.
| ID | Finding | Area | Severity | Broken on |
|---|---|---|---|---|
| F-001 | The canonical tag points at a URL that redirectsRequest the canonical URL directly. It has to return 200 with no hop. | Reach | High | 35% |
| F-002 | No machine readable publish or modified dateView source and search for the year. A date rendered by JavaScript is not a date. | Understand | Medium | 32% |
| F-003 | No author named in the markup or in structured dataLook for a byline in the HTML and an author property in the JSON-LD. Both, not either. | Understand | Medium | 25% |
| F-004 | HTML document heavier than the content justifiesDocument size before images and scripts. Usually inlined state or duplicated component markup. | Speed | Medium | 25% |
| F-005 | Structured data absent, or present and failing to parseExtract every application/ld+json block and parse it. One trailing comma discards the whole block silently. | Understand | Medium | 13% |
| F-006 | A noindex directive on a page that should be indexedCheck the meta robots tag and the X-Robots-Tag header. They are set in different places by different people. | Reach | Critical | not rated |
| F-007 | The server answers 403 or 429 to a declared crawlerFetch with curl and no user agent, then with a browser string, and compare the status codes. | Reach | Critical | not rated |
| F-008 | Content only present after JavaScript runsFetch the raw HTML and count visible words. A browser will always make this look fine. | Read | Critical | not rated |
| F-009 | An AI crawler is disallowed in robots.txtEvaluate robots.txt once per named agent. A bot with its own group ignores the wildcard rules entirely. | Reach | High | not rated |
| F-010 | Sections cannot be lifted as a standalone answerTake one paragraph out of context and read it. If it needs the three above it, it cannot be quoted. | Quote | Low | not rated |
Failure rates are from our own audit of 88 professional marketing sites that returned HTML to our fetcher, published at visibility100x.com/research/geo-agency-audit/. Rows with no rate are checks that did not fail often enough in that sample to have earned one; they are ordered by severity instead.
A PDF cannot be typed into, so a PDF template is a picture of a template. Fill in the Word or Markdown version and export a PDF from that when you send it. That way next month you edit a document rather than rebuilding one.
A Looker Studio template lives inside somebody’s Google account and is shared out of it, which is a hosting commitment rather than a file. We would rather publish nothing than publish a link that quietly stops working when an account changes hands.
Three of the four columns you would otherwise type out are things a tool can answer.
What changed and what it means, the small set of numbers that support it, what was shipped, what is still open and who has it, what you are deliberately not doing, and what happens next. Seven sections. Anything longer is an inventory, and an inventory hands the hard part back to the reader with more work attached.
Yes, and there is no form in front of it. Download the files, edit them, put your own logo on them and send them to a client as your own. No attribution required and no account needed. Gating an artefact behind an email is the obvious growth lever and it is the wrong one.
Yes. Open the workbook in Google Sheets with File then Import, or use the CSV if you only want the findings tracker. The dropdowns are the one thing that do not survive the import cleanly; the values are listed on this page so you can recreate them in a minute with Data validation.
Monthly for most engagements, and the honest caveat belongs in the first one: crawling, indexing and search reporting all lag, so the first genuinely readable read on new work is three to six months in. A monthly cadence is for accountability and for decisions, not because the numbers move monthly.
Because a finding with no decision and no owner is not a finding, it is a sentence. The single most common way an audit fails is that everything gets recorded, nothing gets assigned, and the document is admired once and never opened again. Every row here ends in fix now, fix next, accept, or needs a decision.
By how often the thing is actually broken on real sites, taken from our own audit of 88 professional marketing sites, rather than by category. Importance is an opinion and a failure rate is not. A check can matter enormously and almost never fail, which makes it a poor use of an afternoon.
No, on purpose. A PDF cannot be typed into, so a PDF template is a picture of a template. Fill in the Word or Markdown version and export a PDF from that when you send it, which gives you a document you can update next month rather than one you have to rebuild.
Also no, and for a different reason. A Looker Studio template lives inside somebody's Google account and is shared out of it, which is a hosting commitment rather than a file. We would rather publish nothing than publish a link that quietly rots when an account changes hands.
Free account, no card. Paste your URL and get a real, scored report of your AI and search visibility.
Measuring rankings in