START HERE
Pick one page, explain what the numbers suggest, and choose one thing you can verify or change. A longer task list rarely makes the decision easier.
A weekly analytics session can end with twenty open tabs and nothing changed. Give the session a smaller job: find one page where the evidence points to a useful investigation. This routine is for people maintaining a working website alongside everything else they do.
Start with a question you can finish
Write down the outcome the website is supposed to create. A local service needs suitable enquiries. A shop needs completed orders. A publication may care about useful reading and newsletter subscriptions. That choice changes which pages deserve attention.
Use the last 28 complete days as a practical starting window. Compare an equal preceding period and read any freshness notes. For a seasonal business, check the same season as well. Twenty-eight days is a working routine, not a statistical guarantee.
Separate discovery from the visit
Search Console tells you about appearances and clicks in Google. Analytics tells you about recorded activity after arrival. Google explains why the two sources belong together and why their totals differ in its combined reporting guide.
Write a one-sentence observation for a page. For example: “This guide is being shown for relevant searches, but few searchers choose it.” That statement has a different next step from “Visitors arrive but the enquiry form does not work on a phone.”
Choose a task with an observable finish
A useful task names the page, the suspected obstacle and the check. “Improve SEO” cannot be finished. “Open the delivery guide on mobile and confirm that delivery times appear before the order button” can.
Illustrative example: a fictional furniture guide has 2,400 impressions, 24 clicks and 19 Google organic landing sessions. Its leading query asks whether a desk fits a small room. The article opens with brand history. The first task is to add a verified size answer near the top, then make the search title describe that answer. The figures suggest a place to look. They do not prove the introduction caused the low click rate.
Leave yourself a receipt
Keep the old wording, the new wording, the change date and the report filters in a short note. If you change the title, introduction, navigation and offer together, you may improve the page, but you will struggle to identify which change mattered.
Review the result after enough comparable activity has accumulated. For a quiet site, this may take longer than the next weekly check. Meanwhile, test the page yourself and move to a different task instead of repeatedly rewriting it.
Do this in AbstractMetrics
Open Next actions for your connected website. Read the reason, query evidence and sample limitations on one card. The checklist gives you a starting investigation, and Copy ChatGPT brief carries the evidence into a drafting conversation. Supply the current page copy before asking for exact edits.
If the shortlist is empty, use Pages to inspect an important page manually. An empty shortlist means the current screening rules found nothing qualifying. It is not an all-clear for the website.
What to watch afterwards
Revisit the same page and filters. Compare clicks and the relevant recorded outcome, while checking position, traffic mix and sample size.
Take this to ChatGPT
Help me choose one SEO task for this week. Use my connected website reports if available. Separate observations from possible causes, show the date range and limitations, and ask for current page content before drafting edits.
You choose what to share. Copying does not send anything or connect an account.
Sources and further reading
Official guidance checked on 16 October 2026. Examples and suggested workflows are our own.