CLIENT-01
CLIENT
P0
Evaluating outside SEO advice
A client, or someone on their team, brings us advice from an AI tool that contradicts what we said. The advice usually comes with confidence and a source, and it usually does not have access to the client's own data. We treat it as a live item when it would change one or more things we are doing in the next 30 days.
What This Is
A client, or someone on their team, brings us advice from an AI tool that contradicts what we said. The advice usually comes with confidence and a source, and it usually does not have access to the client’s own data. We treat it as a live item when it would change one or more things we are doing in the next 30 days.
Two facts matter.
First, an AI that was not given the client’s Search Console, crawl, or competitor data is answering from the industry average. It may be right in general and wrong for this site.
Second, dismissing advice out of hand is as bad as accepting it blindly. We answer every piece of outside advice in writing, with data.
Before you check
You have the full text of the advice, not a paraphrase.
You have the exact prompt the person typed, if they will share it.
You have the reasoning or sources the AI returned, if any.
You have our current position in writing, so you can compare.
Diagnostic tree
Run the checks in order. Each check ends in a resolution or the next check.
Check 1. Do you have the source, prompt, and reasoning?
How: ask the person for three things: the tool name, the exact prompt, and the full answer, copied verbatim.
Why: without the prompt you cannot tell whether the AI answered the right question. A vague prompt produces generic advice.
If you have the prompt and the answer, go to Check 2. If you have only a paraphrase, go to Resolution B.
Check 2. Did the AI have our data?
How: read the prompt and the tool setup. Look for any of three inputs in what was fed in: Search Console numbers, crawl data, or competitor data.
Why: if the AI had none of our data, it is giving the industry average. That is not wrong, but it is not specific to this site.
If the AI had our data, go to Check 3. If it had none, go to Resolution C.
Check 3. Does the advice hold against our data?
How: test each claim in the advice, one at a time, against our Search Console numbers, our crawl results, and our competitor data.
Why: the only test that matters is our data. General advice that fails against the client’s own numbers does not get adopted.
If our data supports the advice, go to Resolution A. If our data contradicts it, go to Resolution D.
Resolutions
Resolution | Condition | Action |
|---|---|---|
A. Advice holds | Our data supports it | Adopt it. Thank the person. Update our position and the report |
B. Paraphrase only | No prompt or full answer shared | Ask for the verbatim prompt and answer. Do not act on a paraphrase |
C. Industry average | The AI had no access to our data | Say it is the industry average, not site data. Test it anyway |
D. Advice fails | Our data contradicts it | Respond in writing with our numbers. Do not adopt |
Resolution D in detail
We do not dismiss outside advice. We answer it in writing, and the answer has three parts.
Restate the advice in one or two sentences, so the person knows we read it.
Show the data. For each claim, give the Search Console number, the crawl result, or the competitor data that speaks to it.
Give a verdict. Adopt, reject, or hold. If we reject, say what we would need to see to change our mind.
This is written, not spoken, so it becomes part of the record. The tone is neutral and technical. We never mock the tool or the person.
When to escalate
Escalate when all three are true.
The advice is about buying links, guaranteeing rankings, or publishing unedited AI content.
The person will not accept our written answer.
They still want us to act on it.
These cross our scope line, and we do not do them.
Primary sources
Last verified
·
Owner
Mahesh