SEO agency SOPs are only as useful as their checks, and a check helps only when it is a line someone can pass or fail. A procedure that says make sure the page is optimized has no check in it. A procedure that says confirm the title is under the agreed length and unique within the site has one. Most of the work in writing SOPs is turning the first kind of sentence into the second.
My own reason for believing this is a change log. When two senior reviewers recorded every change they requested for three weeks, about 70% were rule checks and about 30% were judgment, and the SOPs below are built to catch the first group before review. The story is on LinkedIn, and the operating model around it is in SEO agency operations.
The eight SEO agency SOPs worth writing
A small set covers most recurring delivery. Write the one where your last client problem began, then add the others as handoffs become costly.
| SOP | What it settles |
|---|---|
| Client intake and access | What is collected and verified before work starts |
| Scope and backlog setup | How a diagnostic becomes a bounded, prioritized task list |
| Technical audit | The checks run, their order and how findings are rated |
| Content brief | What a brief must specify before a writer starts |
| Pre-publish QA | The checks run before changes go live |
| Post-release verification | What is checked after changes go live |
| Monthly reporting | How the report is assembled, reviewed and sent |
| Scope change and offboarding | How out-of-scope requests and ended contracts are handled |
Every SOP carries the same seven fields: purpose and trigger, a single owner, inputs, numbered steps, checks with a named verifier, the output and where it is filed, and a review date. If a field is blank, the SOP is unfinished. Have someone other than the author run it once before it is published, since they will find the step the author skipped without noticing.
The artifact that matters most: a pre-publish checklist the author fills in
Pre-publish QA is where rule checking piles up, so it is worth building as a gate with evidence in it. The author fills in every line on their own work, and a reviewer only sees work where every line passed. The lines in the structure below are the kinds of rule failures my team's reviewers logged, and the rules in the second column come from a written specification for each deliverable type: the required sections, data sources, naming conventions and any partner-specific requirements.
In my team, template version and layout checks are automated. Every other check is still a checklist run by people, and I treat the rest as candidates for automation once the rules are stable enough to write down.
| Line | Rule, from the specification | Evidence the author attaches |
|---|---|---|
| Template | Current template version in use | Version number as shown in the file |
| Data range | Every figure uses the agreed date range | Range stated beside each chart |
| Disclaimers and labels | Required text present and correctly labelled | Screenshot or quoted text |
| Links | Every link resolves to the intended page | Link-check output |
| Naming | Names follow the written convention | The filename and heading as used |
| Partner or brand rules | Partner-specific requirements applied | Partner sheet line number |
| Indexing | No stray noindex, correct canonical | Directive check result |
Evidence is the part people skip. A tick box invites ticking. A pasted link-check result or a version number cannot be faked without effort, and it lets the reviewer verify in seconds.
The exception log: how recurring requests become rules
The checklist improves through an exception log. Every time a reviewer still sends something back, they write one line: the date, the deliverable, what was requested, whether it was a rule or a judgment call, and whether it was promoted to the checklist. My team runs one, and recurring issues become new rules at a rate of about five or six a month. In my team's log, a chart on the wrong date range and a missing client disclaimer were both rule violations, so both became checklist lines. A judgment call, such as whether a finding is ranked right for this client's goals, stays with the reviewer.
The last column is the point. A request that is a rule and keeps recurring should become a checklist line, and a judgment call stays with the reviewer. As a starting value, promote a rule after it has come back three times, and replace three with whatever your volume supports.
Client access intake: the one check that matters
The intake SOP can be long, but one line carries it. Log in to every system yourself and record the date access was confirmed. Access that was granted but never tested is the usual cause of work that starts on assumption and stalls in week two. Once access works, capture the baseline from it.
Monthly report preparation
Refresh the automated data pull and do not retype numbers between tools. List what shipped, note movement for the areas affected, flag anything that moved without explanation and name what comes next. In my setup the partner reports to their own clients, and my team builds reports on request, so the output has to be something a partner can send as it stands. The check is that every figure in the text matches the data pull and each next step maps to a backlog item.
Rolling out
Pilot each SOP on one live account, keep one current version where the team already works, and make the check part of release so that work without a completed record does not ship. A workflow audit to find missing SOPs is how agency utilization consulting begins. The AI-search versions are in AEO agency SOPs and GEO agency SOPs.
FAQ
What SOPs does an SEO agency need?
A small set covers most delivery: client intake and access, scope and backlog setup, technical audit, content brief, pre-publish QA, post-release verification, monthly reporting, and scope change with offboarding. Start with the one where your most recent client problem began, then add the rest as handoffs become costly.
What should an SEO SOP include?
Every SOP should state its purpose and trigger, a single owner, required inputs, numbered steps, checks with a named verifier, the output and where it is filed, and a review date. If any field is blank, the procedure is incomplete and will be followed inconsistently by different people on different accounts.
What is an exception log and why keep one?
An exception log records each change a reviewer still has to request after work passes the checklist, tagged as a rule or a judgment call. Rules that recur get promoted into the checklist, so reviewers deal with fewer repeated corrections over time and the checklist reflects what actually goes wrong.
Why attach evidence to each checklist line?
A tick box can be ticked without checking anything. Evidence, such as a link-check result, a version number or a quoted disclaimer, shows the check happened and lets the reviewer confirm it in seconds. It turns the checklist from a formality into a gate that work really has to pass.
