TL;DR
If you are putting your name on somebody else’s audit, you need to know whether it is any good before the client reads it, not after. The first test is whether it names a cause the author could be wrong about. Volume is a tell: tools produce fifty-six errors per page, and a good audit reduces that.
If you are putting your name on somebody else’s audit, you need to be able to tell whether it is any good, and before the client reads it, not after.
Here is what I would check, written from the position of somebody who produces these rather than somebody selling you a checklist.
The first test: does it name a cause?
A crawl report lists everything that is true about a site. An audit says which of those true things is the reason the site is not ranking.
That second sentence can be wrong, which is exactly why most deliverables avoid writing it. A list of four hundred issues cannot be wrong. It also cannot be acted on, and your client will not implement it.
So: does the document commit to a cause the author could be embarrassed about in six weeks? If nobody took that risk, nobody did the diagnosis. That distinction is the whole difference between a document worth your name and a tool export with a cover page.
The volume tell
Length is inversely correlated with usefulness here, and there is a reason.
Automated tooling produces enormous volume. WebAIM’s crawl of the top million home pages found an average of 56.1 detected errors per page. On a two-hundred-page site a tool can hand you eleven thousand findings without anybody thinking about any of them.
A deliverable that reflects that volume back at the client is unprocessed. What you want is evidence somebody reduced it: four hundred rows grouped into eight causes, with the reasoning shown.
What it should contain that no tool produces
- Evidence somebody opened the site. On a phone. A crawler does not notice that the home page describes a mood rather than a business.
- Search Console read, not exported. Which queries have impressions and no clicks, and what the position column actually says.
- A competitor comparison. Sometimes the honest finding is that the site is fine and simply outgunned, and that is worth more than forty fixes.
- Three prioritised actions, named, with who does each. Not sixty. Your client can argue with three; nobody argues with sixty, they just do nothing, which looks like agreement and is not.
- What not to bother with. The most valuable page in any audit I write. An audit that recommends everything has made no decisions.
The things that get missed and matter
Two I would specifically look for, because their absence tells you the author was working from a template.
Anything about rendering. Does the audit distinguish what the server sent from what the browser assembled? The 2025 Web Almanac found an invalid element inside the <head> on 10.3% of mobile pages, one in ten, and everything after that break is ignored. A tool reports the canonical as present. It is present; it is just no longer in the head.
Whether pages should exist at all. A thin page can be improved or deleted and both are correct depending on a judgement no tool makes. An audit that only ever says “improve” has not made the judgement.
Practical things, since it carries your branding
No provider fingerprints. Tool watermarks, template headers, the specialist’s phrasing in the executive summary. Check the document properties and the PDF metadata. That is where these survive.
Written for your client’s technical level. A deliverable pitched at a developer landing on a small business owner is unusable, and vice versa. If the specialist has not asked who is reading it, that is a warning about the rest.
You can defend every recommendation. The real test. If you cannot explain why item three is item three, you cannot present it, and you will find that out in the meeting rather than before it.
The short version
- Does it name a cause the author could be wrong about? If not, it is a crawl report.
- Volume is a tell. Tools produce 56 errors per page; a good audit reduces, not reflects.
- Somebody opened the site, on a phone.
- Three prioritised actions with owners. Not sixty.
- A section on what not to do. Usually the most valuable page.
- Check for rendering and for whether pages should exist. Template audits skip both.
- Check the PDF metadata before it goes out with your logo on it.
This is the deliverable I produce for agencies. unbranded, and the client never knows I exist. White label SEO covers how that works in practice.
