Building an AI Search Practice for Enterprise Sites

Challenge

By 2025 the enterprise clients at VML, the WPP agency where I work, all wanted to know whether they showed up when ChatGPT, Google AI Overviews (the AI-written summaries above some Google results), Perplexity, or Gemini answered a question in their category, and whether it changed anything. Nobody had a reliable way to answer. There was no shared way to measure AI citations, no audit method for what makes a page worth citing, and no standards a team of specialists could apply consistently across eight or more accounts, some of them properties with 50,000+ URLs.

Answer Engine Optimization (AEO) is the work of structuring pages so AI systems can retrieve, understand, and cite them. It sits alongside SEO (search engine optimization, the work of getting a site found in search), and on a single page it is a writing task. Across a portfolio of enterprise sites, with engineering, content, and brand teams who are not search specialists, it needs a practice with a shared way to measure, an audit method, tooling that applies the standards consistently, and people trained to run it.

Solution

I built the department’s AI search practice and I run it today.

• Defined a measurement framework built on citation share, AI share of voice, referral traffic from AI platforms, and conversion from AI visitors, each tied to the account’s business goals so AI visibility reports in the same terms as organic search.
• Wrote the audit method and the AI visibility checklist the department now runs on every account. It covers how well the brand and its products are identified, internal linking, structured data (code that labels what a page is about for machines), trust signals, and technical access for the crawlers AI platforms use to fetch pages.
• Built a visibility tool driven by the Model Context Protocol (MCP, a standard that lets AI assistants connect to tools and data), now used department-wide, and orchestrator and sub-agent workflows on Vercel, GitHub, and n8n that automate the analysis and apply SEO, structured data, content, and brand standards at scale.
• Ran before-and-after tests on FAQs, tables, bullets, and other content formats, then folded the results that held up into the department’s content standards.
• Built an AI search training program that more than 30 people have completed, and manage the team of four to six senior managers who deliver the work day to day.

Outcome

On one enterprise program measured over a year, tracked AI citations grew about 60% from baseline to peak, AI-referred traffic grew 174% year over year, and visitors arriving from AI platforms converted at about three times the rate of visitors from traditional organic search.

The automotive inventory program on this site came out of this practice and shows the method applied to one account.

The audits found technical problems too. A post-launch audit of one React site found 1,198 URLs earning Google impressions (appearances in search results) while sending signals that told Google not to index them, because canonical tags and redirects pointed in different directions. The audit came with a roadmap to correct the signals, have the pages recrawled, and confirm the fix. AI systems have to fetch and read a page the way search engines do, so that finding would have capped every AEO result on the site.

The citation and conversion figures are approximate and come from the program’s own measurement, so I report them as “about” figures. The dashboards state the 174% referral figure directly.