An SEO content roadmap for AI Overviews. Six steps, one test.

A working SEO content roadmap starts from what the business needs. You group keywords into clusters, then rank those clusters by difficulty, intent and whether a Google AI Overview appears, with search volume used only to break ties. You write each piece so it can be quoted. Before anything else on the site changes, you test the method on one cluster.

That approach follows the six-step process set out in a Search Engine Journal article published on 2 October 2026. The article is by Corey Morris, President and CEO of the agency Voltage, and describes the roadmap process his agency runs for clients.

Our view: build the roadmap around business intent and terms you can realistically win. Then prove it on a single cluster, measured against a baseline, before you rework the whole site.

What AI Overviews change about planning content

Ranking first is worth less when an AI Overview sits above the results. The article cites an Ahrefs study, which compared 300,000 keywords and found that in December 2025 an AI Overview went with a 58% lower average clickthrough rate for the top-ranking page. Treat it as a sign of direction. It isn't a forecast for your own site.

Overviews aren't rare, either. In the author's logistics roadmap, one appeared on nearly every B2B term examined, including "3pl companies", "3pl fulfillment" and "third-party logistics companies". Those are the commercial searches a logistics business would most want to win.

For anyone planning content, the consequence is practical. You can't assume a page-one ranking brings the traffic it once did, so the roadmap has to record Overviews from the first keyword list onwards. If you add them as a check at the end, you've already chosen your topics on the wrong basis.

The six steps, in order

The process has six steps. Each one feeds the next.

1. Start from the business objective. What the site needs to achieve decides which searches are worth having.
2. Build the keyword universe and run a content gap analysis.
3. Read the SERP, meaning the search results page for each term.
4. Cluster and group the keywords.
5. Prioritise.
6. Hand off to the brief and write for AI visibility.

The gap analysis is easy to rush. In Ahrefs Content Gap, you enter your own domain and then add competitors: at least two to four, and up to 10. Choose competitors that sell what you sell to the same kind of customer. A large site in a different line of work will distort the list.

Reading the SERP means looking at the live results for each term yourself, as well as the numbers in a tool. That's where you see whether an AI Overview sits at the top, and that observation carries through every later step.

For agencies, the value is repeatability. The same six steps apply to each client, so a roadmap for a dental practice and one for a trade body follow the same order and produce the same kind of sheet. You explain the method to a client once, then reuse it.

How to organise keywords into a working sheet

Everything lives in one sheet with seven columns: Cluster, Group, Keyword, KD (keyword difficulty), Volume, Intent and SERP Features. The structure is nested as Cluster > Group > Keyword. A cluster is a broad topic; a group is a set of closely related searches inside it. The keyword is the individual search term.

The author's example comes from logistics. A Foundation cluster holds a 3PL group with "3pl companies", "3pl services" and "3pl warehouse". A separate Compliance cluster holds an Amazon FBA Preparation group with "amazon 3pl" and "amazon fba prep center". Both involve third-party logistics, but the second is tied to one specific Amazon task. Keeping them apart stops a single page trying to answer two different needs.

The SERP Features column is where you record whether an AI Overview shows for each keyword. Fill it in from the live results and note the date you checked. Results change, and your baseline later depends on knowing what you saw and when.

How to decide what to write first

This is where volume tempts people. In the author's example, the head term "3pl" has a KD of 47 and 29,000 searches a month. It looks like the obvious first target.

The team didn't start there. Here are the three terms side by side:

• "3pl": KD 47, 29,000 searches a month.
• "3pl fulfillment": KD 11, 2,600 searches a month.
• "3pl companies": KD 18, 5,000 searches a month, with more clearly defined B2B intent.

They put "3pl fulfillment" and "3pl companies" into the Foundation cluster instead of "3pl". Both are far easier to rank for, and "3pl companies" is the search of a business looking for a supplier. They also moved the Amazon FBA prep cluster up the list, because its intent is closer to buying.

No tool scores this for you. The author says prioritisation combines difficulty, potential, intent and the reality of whether an AI Overview appears, with volume as the last tiebreaker. There's no formula. It's a judgement, so write the reasoning next to each cluster in the sheet. Anyone approving the plan can then see why one cluster came before another.

The common mistake is chasing the biggest volume figure first. A hard term with an Overview above it is difficult to win, and even a top ranking gives up clicks to the summary. Two easier terms with clear buying intent make a better first test.

How to write pieces that AI Overviews can cite

The last step hands each cluster to a brief. The article gives four rules for writing that can be cited, and each one can be checked in a draft before it's approved.

• Open each section with a direct answer of two to three sentences. A reader, or a summary, should get the point without reading further.
• Use headings that mirror how people search. For "3pl companies", a heading such as "How to choose a 3PL company" is closer to the search than a clever label.
• Add schema or structured data to the page.
• Link internally back to the hub page for the cluster.

Put these in the brief as plain instructions. The approver then has a short checklist: does each section answer first, do the headings match real searches, is the structured data there, and does the piece link to its hub? If a draft fails one, it goes back before it goes live.

This suits small teams. A brief with clear rules is quicker to check, whoever writes the piece.

How to tell whether it's working

Before you change anything, record a baseline. Note the current SERP features for each target keyword, including which ones show an AI Overview, and export your current click data. Date both.

Then test the method on one cluster only. Build that cluster's pieces the new way and compare them with content built the way you worked before. The author recommends exactly this, and it's sound advice: the article reports no results, timings or costs for the method. Your own comparison is the only evidence you'll have of what the roadmap delivers on your site.

For an agency, one cluster on one client site is a contained test that's easy to explain and report on. For a small business or member organisation, it means you aren't rewriting dozens of pages on the strength of a method you haven't yet seen work.

So start with one. Choose the cluster closest to what your business sells, check its SERP features, record the baseline and write the first piece from the brief.

If you want approved, voice-matched articles for your own site, see how the Desk works.