How to track rankings and AI visibility

Why tracking needs care

You can’t improve what you don’t measure, but search is easy to measure badly. Your position for a search is not one fixed number. It changes with the searcher’s country and town, their device, the language, what Google is testing that week and, for AI answers, with each time the question is asked.

Good tracking accepts this. It measures the same things, the same way, on a schedule, and judges trends over weeks rather than single readings. This guide shows how to do that for Google rankings and for AI answers.

Four sources, four different questions

Start with Search Console

Search Console is the only source with Google’s own counts. Its average position is the average of your highest position each time you were shown, so it blends many places, devices and days into one figure. That makes it excellent for spotting direction and poor for answering “where am I right now for this phrase?”. Our guide to reading Search Console data explains its numbers in full.

Add rank checks for the searches that matter

A rank check runs a search the same way every time and records where your pages appear. It is a snapshot, not the whole truth, but because the method never changes, two snapshots can be compared fairly.

1. Choose the keywords. Take them from your keyword master list: the searches that bring customers, plus a few you are working towards. Our guide to building a keyword master list shows how. Twenty well-chosen keywords tell you more than five hundred random ones.
2. Fix the settings. Same country, same language, same device every time. Phones are a sensible default, because most searches happen on them and Google ranks the phone version of pages.
3. Check on a schedule. Weekly is enough for most sites. Daily checks show mostly noise; positions move a little every day for reasons that have nothing to do with you.
4. Record which page ranks. If the page that ranks for a keyword keeps switching between two of yours, they may be competing. See mapping keywords to pages.
5. Track a few competitors too. Your own position means more next to theirs. A drop that every competitor shares is usually a change in the results, not a fault on your site.

If your customers are local

For searches with a place in mind, such as “florist near me” or “bakery Leeds”, results change from street to street. A rank check for the whole country can say you are 12th while a customer two streets away sees you in the map at the top. For these searches, also watch your Google Business Profile’s own figures (searches, calls, direction requests), and treat country-level positions as a rough guide only.

Turn many positions into one trend

A list of fifty positions is hard to read. Two simple summaries help. Count how many of your keywords are on the first page of results this week against four weeks ago. And weight that by how often each keyword is searched, so a move on a busy search counts for more than a move on a rare one. Both are easy to keep in a spreadsheet, and both show direction better than any single keyword.

Whatever you choose, keep the same method for months. Changing the keyword list or the settings every few weeks makes every comparison meaningless.

Track how AI answers talk about you

People now ask AI tools questions they used to type into a search box: “which garden shops deliver plants in Leeds?”, “is example.com reliable?”. There is no Search Console for most of these tools, so you measure by asking.

1. Write a fixed set of questions. Use the questions a real customer would ask before buying: best options, comparisons, how to choose, and questions about your business by name. Keep the set stable so weeks can be compared.
2. Ask them on the tools your customers use. For example ChatGPT, Gemini, Claude, Perplexity and Google’s own AI answers. Use the same settings each time, such as signed out or a fresh conversation.
3. Record three things per answer. Were you named? Was one of your pages cited as a source? Which competitors were named instead?
4. Count across many questions. AI answers vary from one asking to the next. A single answer means little; being named in 6 of 30 questions this month against 3 of 30 last month is a real signal.

Our guide to brand and citation signals explains what tends to make AI tools name and cite a business.

Reading the numbers without being misled

A simple reporting routine

• Weekly, five minutes: scan rank changes for your main keywords and any new AI mentions. Note anything that moved a lot.
• Monthly, half an hour: compare 28 days with the 28 before in Search Console, list keywords that rose and fell, and count AI mentions against last month.
• Every quarter: review the keyword list and the AI questions themselves. Drop what no longer matters and add new products or services.

Write each month’s numbers and the changes you made in one place. The point of tracking is the next decision, and our guide to fixing performance gaps shows how to turn these numbers into work.