Run Your Monthly AI SEO Review in Four Questions

The insight: your monthly SEO meeting should produce one decision, not forty charts

Most law firm marketing reviews follow the same script. Someone shares a dashboard. Traffic is up or down. Rankings moved a little. Everyone nods. The meeting ends. Nothing changes until the next meeting, when the same numbers move a little again.

Here is a better structure. Your monthly AI SEO review should answer exactly four questions:

• What changed?
• Why did it change?
• What did it affect?
• Who owns the next move?

That is the whole agenda. Four questions, in that order, and the meeting does not end until the fourth one has a name attached to it. Not "marketing." Not "the agency." A person, and a date.

The reason this works is that it forces causality. A dashboard tells you a number moved. It does not tell you whether the movement came from a Google update, a page you published, a competitor's new content, a technical regression, or nothing at all. Question two is where most firms discover they have no idea why anything is happening — which is itself a useful finding, because it tells you what to instrument before next month.

Add AI answers to the agenda

The four questions used to apply only to blue links. They don't anymore. Your monthly review needs two more data points on the table:

Verified AI citations

When an AI assistant answers a question in your practice area and your market, does it cite your pages? Which pages? Note the actual citations you can verify — not an impression, not a guess. If you cannot verify them, that is question two territory: you need a way to check before you can manage it.

Brand mentions in AI answers

Separate from citations, does the AI answer name your firm at all? A citation is a link. A mention is a recommendation. Firms often have one without the other, and the gap tells you something about whether your content is being read as an authority or just as a source of facts.

Track both month over month with the same four questions. What changed in your citations? Why? What did it affect — did intake volume move, or did nothing happen downstream? Who owns fixing it?

The one rule that belongs in every review: schema must match the visible page

Here is the source fact to build your standing action item around. Google requires structured data to represent the visible page accurately, and warns against hidden, irrelevant, misleading, or fake-review markup.

Read that carefully, because it is stricter than most firms treat it. It is not "schema should be roughly related to the page." It is: the structured data must represent what a human being actually sees when they load the URL. Markup describing content that is hidden, markup that is irrelevant to the page, markup that misleads, and fabricated review markup are all called out explicitly.

This matters more in AI search than it ever did in classic search. Structured data is one of the clearest ways a machine understands what a page is, who it is for, where it applies, and who published it. When your schema is accurate, you are handing an AI system a clean description of your page. When it is inflated, you are handing it a description that contradicts the page — and contradiction is exactly what these systems are built to discount.

Inflated schema is not a shortcut. It is a contradiction between what your page says and what your markup claims, and contradiction is precisely what modern search systems are built to discount.

Law firm sites accumulate this problem quietly. Review markup gets added from a plugin that pulls from somewhere nobody remembers. Attorney schema lists credentials that are no longer on the bio page. Service pages carry markup for services that were removed two redesigns ago. FAQ markup describes questions that got deleted from the visible copy. None of it was malicious. All of it now fails the accuracy standard.

How to close the meeting

End every monthly review by committing to one action. Not a list. One. This month, make it this:

Make every schema property match visible content on the page.

Here is a practical order of operations for the person who owns it:

• Start with review markup. It carries the most risk and gets the most scrutiny. Confirm every review represented in markup is a real review that appears on the page.
• Audit templated pages next. Anything generated from a template repeats the same error across every URL using it. Fixing the template fixes hundreds of pages at once.
• Check attorney and organization markup. Credentials, practice areas, service areas, addresses — do they match what is actually printed on the page?
• Remove, don't rewrite, irrelevant markup. If a property does not describe something a visitor can see, delete it. Missing schema is a smaller problem than wrong schema.
• Document what you changed so next month's "what changed / why" questions have an honest answer.

Where this gets hard at scale

The accuracy rule is simple on a fifteen-page site. It is a different problem when your footprint is large — every practice area across every city you serve, each with its own page, its own schema, its own internal links. That is the coverage most established firms want and almost none of them build by hand, because building it by hand means maintaining thousands of individually correct schema blocks forever.

The only way that scale stays accurate is if the schema and the visible content are generated from the same source of truth. If the page says you handle a specific practice area in a specific city, the markup says exactly that — because both came from the same record. When they are maintained separately, they drift, and drift is how you end up with markup that describes content that is not there.

That is the principle behind how we approach programmatic local coverage: page inventory at scale, with schema, internal linking and AI-search optimisation built in rather than bolted on afterward. You can see how we approach AI SEO if you want the longer version.

Your next step

Put the four questions on your calendar as a recurring 30-minute meeting. Add verified AI citations and brand mentions to the standing agenda. And before the next one, have someone pull up your five highest-value pages, view the source, and compare every schema property to what is actually visible on screen.

If they match, you have a clean baseline. If they don't, you just found your one action for this month — and you will know exactly who owns it.