The top-ranking page loses roughly 58% of its clicks when a Google AI Overview is present, according to Ahrefs research cited in a new step-by-step framework for building SEO content roadmaps in the AI search era.

Corey Morris, whose agency team runs the process for clients, published the six-step roadmap on Search Engine Journal. His starting premise: AI makes content fast and cheap, but the "good" corner of the project management triangle — good, fast, cheap, pick two — still depends on human discipline.

"In AI Overview SERPs, we can experience a lot of waste even if we are visible if we're not disciplined in our research," Morris writes. His process targets AI Overviews specifically, which he calls the biggest current intersection between SEO and Google, though he says the same discipline extends to LLM visibility in tools like ChatGPT and Perplexity.

Step 1: Anchor to a business objective

Every keyword, topic, or concept must map to a trackable outcome — a lead, ecommerce revenue, or another metric meaningful to the organization. Morris argues that visibility or traffic alone is not a deep enough goal, except for media companies monetizing pageviews through ads.

His example: a relationship-driven third-party logistics provider does not want leads hunting the lowest shipping cost. It wants businesses with complex needs, and success means content investment that drives form conversions, booked sales, and revenue over the relationship. Teams that cannot define and quantify the goal risk producing content that never meets stakeholders' ROI expectations.

Step 2: Build the keyword universe

Morris's team uses Ahrefs for the workflow. The process: run Content Gap analysis against two to four search competitors (up to 10 allowed), filter by keyword difficulty and volume, set ranking position ranges, and export the output for clustering later.

Competitor selection is a judgment call, he stresses: pick "search" competitors you actually face in the SERP, not the product competitors your sales and product teams discuss internally. The raw output is also not a to-do list — a large share will be branded, off-intent, or irrelevant.

The 3PL example makes the trade-off concrete. The head term "3pl" carries a keyword difficulty of 47 and 29,000 monthly searches — tempting, but broad and hard. "3pl fulfillment" sits at KD 11 with 2,600 monthly searches, and "3pl companies" at KD 18 with 5,000, both with sharper B2B intent. "There are sweet spots to try to hit before you take on the top industry term," Morris writes.

Step 3: Read the SERP, not just the keyword

Volume and KD indicate whether a page can rank. They do not say what the ranking is worth. In an AI search SERP, the first organic spot can sit below an AI Overview, ads, and other features — so the same position can deliver a fraction of its former clicks.

The Ahrefs workflow filters keyword ideas by SERP features to isolate terms triggering AI Overviews, then checks Site Explorer to see where a site's pages are already cited inside them. Each keyword's SERP features get recorded on the roadmap.

An AI Overview presence is not an automatic skip. For each term, teams decide whether it changes targeting or just how they write for it — writing to be cited on informational terms, or competing for the organic result below.

In the logistics roadmap, an AI Overview appeared on nearly every B2B term the team examined: "3pl companies," "3pl fulfillment," "third-party logistics companies." That finding reshaped the entire content plan, since visibility on those terms now means citation inside the AI Overview.

Step 4: Cluster by business meaning, not statistics

"A keyword list isn't a plan," Morris writes. Authority in AI systems comes from clusters and topics, not one-off pages chasing single keywords. Ahrefs clusters keywords by Parent Topic — terms sharing the same top-ranking page — or by shared terms, and each cluster maps to a destination: hub page, service page, or article.

The tool clusters by ranking overlap, but the human call is clustering by business meaning. In the 3PL roadmap, core service terms rolled into a Foundation > 3PL cluster ("3pl companies," "3pl services," "3pl warehouse"), while "amazon 3pl" and "amazon fba prep center" formed a separate Compliance > Amazon FBA Preparation cluster. The FBA cluster is bottom-of-funnel, a specific need, not a variant of the general terms — a distinction the tool missed.

Step 5: Prioritize on four factors

Prioritization weighs difficulty, potential, intent, and AI Overview reality. No tool scores this combination, Morris says — it is the human layer. Clusters get scored against the Step 1 objective first, with proximity to lead or revenue outranking volume, then tiered into a Foundation set built first.

In the 3PL plan, the team passed over the broad, AI Overview-dominated term "3pl" and prioritized "3pl fulfillment" and "3pl companies" into Foundation. It pushed the Amazon FBA prep cluster up the list because search intent sits closer to buying, despite fewer monthly searches. "Volume was the last tiebreaker here and not the first filter," Morris writes.

Step 6: Write to be cited

Each prioritized cluster becomes a brief: primary keyword, supporting terms, intent stage, SERP features, target page, and angle. The writing guidance is built for AI extraction: lead each section with a two- to three-sentence direct answer, use headings that mirror real search phrasing, and include specifics AI cannot fake — real expertise, concrete detail, named entities, data. Schema and internal links from supporting pieces to the hub complete the technical layer.

AI can accelerate briefs and first drafts, but the credibility comes from human examples, expertise, and editorial judgment. That, Morris argues, is what protects against competitors shipping "fast" and "cheap" AI-only content that cannot replicate genuine expertise.

Baseline first, then test

Morris closes with two actions. Capture current SERP feature and click data as a baseline before AI features shift again, because the click math already differs from two years ago and will keep changing. Then test the method on one cluster rather than rebuilding the whole content operation — compare its performance against a piece built the old way and let the result guide adoption.

His warning for teams that skip the discipline: generating visibility without a mapped business outcome is a newer way to waste SEO and AI optimization budgets, and a harder one to catch, because impressions and rankings can still look healthy in reporting.