The weekly routine I run in Semrush

A good tool without a routine piles up data nobody reads. The routine is what turns data into decisions every week.

Disclosure: I am a Semrush Ambassador, and this article describes Semrush features. The working criteria described here hold with any equivalent tool, and the passages describing the product are marked with source and date checked.

The difference between people who get value out of an SEO platform and people who just pay the subscription usually comes down to a fixed routine, in a defined order, that ends in a decision. Without it, the platform becomes a very expensive way to feel informed.

Mine has four blocks and fits in ninety minutes per account, give or take depending on the size of the project.

Block 1. Technical health, fifteen minutes

I open Site Audit and look only at what changed since the last audit. A new error matters more than an old one, because a new error has a recent, traceable cause. I look for pattern changes by template, paying attention to crawl blocks and pages that dropped out of the index.

Block 2. Ranking movement, twenty minutes

In Position Tracking I look at drops before gains. A drop concentrated in one group of pages points to a structural cause. A scattered drop points to a change in the search surface or an algorithm update, and calls for a different kind of investigation.

Block 3. Territory and competition, thirty minutes

Keyword Gap and Backlink Gap against the three competitors fighting for the same results. Here I look for movement over time, with a single-day snapshot serving only as reference. A competitor that gained ten links from the same type of source in two weeks is executing a plan, and their plan is useful information.

Block 4. AI visibility, twenty-five minutes

This block of the routine depends on a specific set of features, so it is worth logging what it covers, with the date checked.

In this block I compare the brand’s presence with the previous week and log any change in its description, because a description that changes without warning usually means a new source entered the set the system reads. The full design of this measurement is in how to track AI visibility.

What goes to the monthly routine

Some readings make no sense in a one-week window and go to a separate monthly routine: comparing the full audit with the previous month, reviewing the prompt set, checking the external sources that describe the brand, and reading referral traffic from AI platforms, which swings too much to be read week by week.

Keeping the two routines apart avoids reacting to noise as if it were a trend, and keeps the weekly one at ninety minutes.

What closes the routine

Each block ends with one written line: what changed, what it means and what the action is. Without that line, the routine turns into dashboard watching. With it, the following week starts with a short list of decisions already made.