MarketingContent & SEO

Your team plans more pillar pages for GEO, and AI answers cite a partner's listicle instead

Generative engine optimisation changes three things in how a content team runs: what it plans around, what it produces, and what it reports. This playbook covers how to move from keyword-to-page production to prompt families, off-site evidence and citation share, using a decision rule you can apply this quarter.

Workers add marble pillars to a grand temple while a giant colored pin lands on a small neighbouring tent.

Semrush's guide to pillar pages still teaches the classic architecture. From your pillar page, you would add links to cluster pages. These would dive deeper into the subtopics mentioned on your pillar page. The guide's tooling advice tells you to start by entering one or more seed keywords representing your website's core topics, and the tool returns each pillar page with its list of subpages. Semrush sells SEO tools, so the guide is also a product walkthrough. The architecture still works for getting indexed and ranked. The trouble starts after the page ranks. Seer Interactive studied 53 brands and 2.43 billion impressions, and found that when an AI Overview appears, pages cited in it get more clicks than pages on that same results page that aren't cited. Its numbers: roughly 3.3% CTR with no AI Overview, 2.1% when the page is cited, and 0.9% when it isn't. Across the whole web, SparkToro's June 9, 2026 study, built on Similarweb's U.S. desktop and mobile web panel for January through April 2026, found 68.01% of Google searches ended without a click.

Most content teams have responded to GEO by changing how they write: shorter answers, more FAQ blocks, AI-assisted drafts. That is the smallest part of the change. The bigger shift is in the operating model. Teams now plan around prompt families instead of keywords. The thing they produce is an answer that gets cited, and that answer can live on a site they do not own. They report citation share and the pipeline it feeds, with sessions as a secondary number. A team that only retools its writing will turn out better-formatted pillar pages for a channel that hands much of its credit to other people's content.

How should a content team reorganise for GEO?

Reorganise around prompt families, give each family a named owner, split the work between owned answers and off-site evidence, and report citation share next to traffic. The steps below are in the order a CMO can run them, starting with a decision rule that settles most planning arguments.

A whiteboard rule: follow the citation

Before you commission any asset, run its prompt family through the AI engines your buyers use and see who gets cited. Then apply this rule:

  • Your own page is cited: refresh and extend it (owned work).
  • Competitors or third parties are cited: earn a place in those sources (off-site work).
  • Nobody credible is cited, or the answer is generic: publish original data or expert material (evidence work).

On the whiteboard it reads: Owned cited, refresh. Others cited, earn. Nobody cited, publish proof. The rule moves budget toward wherever the answer is actually being sourced. Without it, a content calendar becomes a list of pages your team happens to know how to make. It also answers the format question. One practitioner guide on prompt gaps says to decide the format from the evidence, and notes that if comparisons and user reviews dominate, another explainer will not help.

Step 1: swap the keyword map for a prompt portfolio

Replace the keyword spreadsheet with a list of prompts grouped by buying stage. Semrush's latest methodology note (again, from a tool vendor) splits prompts into two groups and tags them separately from Core prompts. Each tag has its own AI Visibility score. The same note admits the limit: Prompt Tracking only checks the prompts in your campaigns, and no tool has a log of every conversation buyers have with AI. Scale is a budget decision too. The entry AI Visibility Toolkit plan covers 25 prompts, so a full portfolio means paying for a higher tier.

Do not assume your SEO gap analysis covers this. HubSpot, which sells its own AEO tool, warns that an SEO keyword gap and an AEO keywords gap don't tell the same story, and closing one doesn't close the other. In practice, the prompts that cost the most to miss are comparisons ("X vs Y for mid-market") and constraint prompts about integrations, compliance or company size, where a specific answer beats a general one.

Step 2: keep the pillar, change its brief

Keep the hub-and-spoke structure. It still decides what gets indexed and ranked, and Google's AI Overviews draw on that index. What changes is the brief. Each pillar section should open with a direct answer, and each cluster page should map to a prompt family rather than a keyword with enough volume.

That second change matters more than it sounds. Semrush's AI Overviews study (vendor data) found nearly 60% of keywords triggering AI Overviews had 100 or fewer searches a month. If your brief template has a volume floor of 500 searches, it filters out most of the questions AI answers are built on. Another Semrush study found that ChatGPT search cites pages ranking in position 21 or lower almost 90% of the time. Ranking first for a few head terms counts for less than covering many specific questions well. For the structural side, the team should revisit its topic cluster architecture with prompt families as the organising unit.

Step 3: stand up an off-site evidence desk

This is where the operating model breaks with the past. Semrush's affiliate team reports from its own program that roughly 1 in 5 partnered articles about Semrush is cited in tracked AI answers, and that listicles and comparisons are the affiliate content most often cited in AI answers. The data comes from a vendor describing its own partners, so treat it as one case and not a benchmark. The mechanism is still clear. Semrush argues that AI systems build their answers largely from third-party content, and affiliates publish a lot of it.

The publishers that answer engines rely on are also organising. On November 12, 2026, a group of more than 3000 media brands, content owners and publishing platforms launched the Internet Licensing Collective, and Yahoo Scout, a new AI answers tool from Yahoo, is confirmed as ILC's first commercial licensee. Mather Economics, a publishing analytics firm, estimates that publisher payments under collective AI licensing could amount to 15 to 35 per cent of AI-company revenue. For a CMO, this means the outlets AI engines cite are becoming paid, permissioned suppliers to those engines. Coverage in those outlets is a GEO input, yet it usually sits in the PR budget and is reported against a different goal.

The move: put PR, affiliate, partner marketing and analyst relations on the same prompt portfolio as content. One person, the evidence desk lead, decides which third-party sources to pursue for each prompt family where "others cited" applies.

Step 4: rebuild the scoreboard around citations

Traffic still matters, but it no longer shows whether the content program is working. Here is a hypothetical worked example that uses only Seer Interactive's published rates. Take a prompt family that earns 100,000 Google impressions a month:

  • With no AI Overview at 3.3% CTR: 3,300 clicks.
  • With an AI Overview that cites you at 2.1%: 2,100 clicks.
  • With an AI Overview that skips you at 0.9%: 900 clicks.

The AI Overview takes away 2,400 clicks if you are not cited and 1,200 if you are. So the citation wins back half the clicks the Overview removes. Ranking alone wins back none of them. That is why citation share by prompt family belongs on the dashboard.

Conversion quality adds to the case. In an Ahrefs case compiled by The Digital Bloom (Ahrefs sells SEO tools, so read it as the vendor's own data), AI search traffic, just 0.5% of total visitors, produced 12.1% of signups over a 30-day measurement window. To size the overall risk, a Driven Metrics review of eight studies puts the typical AI Overview click-through rate penalty at about 40%, with a median estimate of 43.3%.

The new monthly report has four lines: citation share per prompt family, clicks to cited pages, AI-referred conversions, and how many prompt families sit in each bucket of the follow-the-citation rule.

Step 5: redraw roles and cadence

Four roles cover most mid-size teams. A prompt-family owner (often the former SEO lead) is accountable for citation share. An evidence editor commissions original data and expert review. An evidence desk lead handles off-site sources. An analyst keeps the tracking method fixed. Writers sit under the family owners, not under a channel. If you are rebuilding the org chart, the work of redesigning roles around revenue outcomes applies directly.

Cadence matters because AI engines favour recent content. The Digital Bloom reports that 65% of AI bot hits target content < 1 year old, so pillar pages need a refresh date in the plan rather than "evergreen" status. Start tracking early: in Semrush's tool, Prompt Tracking starts collecting the day you create your campaign, with no backfill. Every week of delay is a week of baseline you never get back.

Common GEO operating model mistakes and how to avoid them

Treating prompt count as coverage. Adding more tracked prompts feels like progress. Semrush's own methodology piece warns that tracking more prompts can still leave blind spots. Fix the mix of core, exploratory and verification prompts before you grow the list.

Building a separate GEO squad. A standalone team chasing citations while the SEO team chases rankings will fight over the same pages. Prompt-family owners should own both outcomes for their family.

Reading the CTR rebound as a reprieve. Seer's recovery to 2.4% is real, but it remains well below the roughly 3.3% CTR on searches where no AI Overview appears at all, so the underlying click loss hasn't reversed. Budgets planned on a return to old traffic levels will miss.

Counting mentions and ignoring framing. As Search Engine Land puts it, visibility without favorable context isn't the same as brand strength. An AI answer that names you as the expensive option in a comparison prompt is a cost, not a win. Log sentiment for each prompt family.

Mass-rewriting the archive with AI. Even HubSpot is now publishing guidance on using AI to write blog posts that aren't slop. Run rewrites only on families in the "owned cited" bucket, where you already have the source position.

Comparing visibility scores across tools. Semrush, Ahrefs, HubSpot and Scrunch each publish comparisons of their rivals' AI visibility tracking, and each defines "visibility" its own way. Choose one tool, write down the method, and keep it fixed for at least two quarters. In Semrush, each tracking campaign follows one AI platform, so budget for one campaign per engine your buyers use.

What to do on monday

  • Write 30 prompts for your highest-revenue product line, phrased the way buyers ask, tagged by buying stage.
  • Run them through ChatGPT, Gemini and Google AI Mode, and sort each into owned cited, others cited or nobody cited.
  • Remove the search-volume minimum from your content brief template.
  • Ask your affiliate and PR leads for the list of comparison articles and listicles that mention you, and check which ones show up in AI answers.
  • Set up prompt tracking this week so the baseline starts now.
  • Add citation share per prompt family as the first line of next month's content report, above sessions.

Your pillar pages are still worth keeping, since they are what gets your brand into the index. What GEO changes is who owns the answer and where that answer lives. Give each prompt family an owner this month, and expect some of those owners to spend more time on other people's websites than on your own.

Frequently asked questions

What is the difference between SEO and GEO for a content team?

SEO plans content around keywords and measures rankings and clicks, while GEO plans around prompt families and measures how often AI answers cite the brand. HubSpot warns that closing an SEO keyword gap does not close the AEO gap, so teams need separate tracking even when they keep the same pillar and cluster pages.

Do pillar pages still work for AI search?

Pillar pages still matter because Google's AI Overviews draw on indexed, well-linked pages. Their brief needs to change: each section should open with a direct answer, and cluster pages should map to prompt families. Semrush found nearly 60% of keywords that trigger AI Overviews have 100 or fewer monthly searches, so volume-based topic selection misses many of them.

How much traffic does being cited in an AI Overview save?

Seer Interactive measured about 3.3% CTR with no AI Overview, 2.1% when a page is cited and 0.9% when it is not. On 100,000 impressions that works out to 3,300, 2,100 and 900 clicks, so a citation wins back half the clicks the AI Overview removes.

Why do affiliate and PR teams matter for generative engine optimization?

AI answers rely heavily on third-party content such as comparisons, reviews and listicles. In Semrush's own affiliate program, roughly one in five partnered articles is cited in tracked AI answers. A content team that only works on its own site leaves the prompt families where other sources get cited without an owner.

Go deeper

The lessons that take this article further, free to read.

  1. 1SEO & content marketing: frameworks & methodologyDemand generation
  2. 2CMO playbook & advanced tactics: SEO & content marketingDemand generation
  3. 3SEO & content marketing: real-world applicationDemand generation
  4. 4AI & ML in marketing: real-world applicationMarTech & data
  5. 5CMO playbook & advanced tactics: building the team that drives revenueLeadership & organization

Sources

Finished reading?

Validate your read to earn XP and feed your radar.