How to Get Your Site Cited in Google's AI Answers

What a theme rebuild did to one national brand's AI visibility, and why fixing pages one at a time will not get you there

The short answer

Appearing inside Google's AI-generated answers is mostly a structural problem, not a content problem. An AI answer is assembled by a machine reading your pages, so whatever makes a page hard for a machine to parse, tangled markup, inconsistent headings, missing or wrong structured data, slow and shifting layouts, keeps you out of the answer no matter how good the writing is. Alan Morgan Group rebuilt the WordPress theme underneath a national multi-unit residential home services company rather than fixing pages one at a time, because the theme governs every page type at once. Impressions in Google's generative AI features went from about 154,000 in one three-month period to about 522,000 in the next, roughly 3.4 times the previous rate, while organic search users grew 18.4% over a comparable stretch. The lesson is that on a large site, AI visibility is won at the template level.

Key figures

• Impressions in Google's generative AI features rose from about 154,000 to about 522,000 between the most recent three months and the three months before them, an increase of about 239%. Source: Google Search Console, Performance on Search results, Generative AI features report (beta), three-month comparison view, retrieved 17 September 2026.
• That is about 3.4 times the previous rate, or roughly 368,000 additional impressions.
• Organic search users rose 18.4%, from 69,700 to 82,500, comparing January 1 through April 30, 2026 against May 1 through September 16, 2026.
• Organic search clicks rose 18% over the most recent three months, to 52,600, while impressions held flat at 7.46 million, down 1%. Source: Google Search Console, Insights, three-month view, retrieved 17 September 2026. Two independent measurement systems, one analytics platform and one search console, agree within half a percentage point.
• More than 220 individual improvements were completed across 9 workstreams.
• 1 theme rebuild, live in mid-July 2026, touched every page type on the site: location pages, state and city templates, service-area pages, blog posts and archives, author and category pages, search results, and paid landing pages.

What changed about search, and why the old checklist stopped working

For twenty years, the job was to get a link placed in a list. Somebody typed a query, Google returned ten blue links, and the work was about occupying one of them.

An AI answer is a different object. Google now assembles a written response by reading across multiple sources and synthesizing them, then names some of those sources. Being one of the named sources is a different achievement from ranking tenth, and it is not won by the same tactics. Nobody clicks through a list to find you. Either the machine could read your page well enough to use it, or it could not.

That shift moves the center of gravity from the page to the plumbing. A well-written article on a page with broken heading structure, missing structured data and a layout that shifts while it loads is a page the machine will struggle to use, and it will quietly use a competitor's instead. No error message, no ranking drop, no obvious symptom. You simply do not appear.

Alan Morgan Group works on this for multi-location and national operators, where the problem compounds because the same structural flaw is replicated across thousands of URLs.

Why does a site get skipped by AI answers?

Four structural causes account for most of it, and none of them are about the quality of the writing.

Markup the machine has to guess at. Heading levels used for visual size rather than hierarchy, content buried in nested containers, important text rendered by scripts. A human reader sees a clean page. A parser sees ambiguity, and ambiguity gets skipped in favor of a source that is unambiguous.

Structured data that is missing, partial, or contradicting the page. Structured data is the explicit version of what a page is about. Without it the machine infers, and inference loses to a competitor who simply stated it. Worse than missing is structured data that disagrees with the visible content, which reads as a reliability problem.

Performance that interrupts the read. Layout shifts, heavy scripts and slow font loading degrade how reliably a page can be rendered and evaluated. This is the same work as Core Web Vitals, which is why the performance and AI-visibility agendas turn out to be the same agenda.

Inconsistency across page types. A site where location pages, blog posts and service pages each express structure differently offers no repeatable pattern. Consistency across templates is itself a signal, and it is impossible to achieve page by page on a site with thousands of URLs.

Why fix the theme instead of the pages?

Because on a large site, the theme is the only place where a fix applies everywhere at once.

A national operator with locations, service areas, a blog and paid landing pages has thousands of URLs generated from a few dozen templates. A structural flaw in a template is a structural flaw in every page that template produces. Fixing pages one at a time means fixing the same defect thousands of times, and the site will outgrow the effort while it is underway.

Alan Morgan Group rebuilt the theme for this engagement rather than applying isolated fixes. The rebuild covered the global header and navigation, the footer, location pages, state and city templates, service-area pages, blog posts and archives, author pages, category and tag archives, search results, paid landing pages, review and testimonial components, FAQ components, forms, accordions, carousels, responsive layouts, mobile navigation, typography and image handling.

The point was not a visual refresh. It was that cleaner markup, correct heading structure, consistent canonical handling, proper structured data and improved asset loading could be established once, in the framework, and inherited by every page built on it from then on. That is what makes the improvement hold as the site keeps growing, and it is why the payoff shows up as a step change rather than a slow grind.

More than 220 individual improvements were completed across nine workstreams: theme and template redevelopment, technical SEO remediation, local SEO and location infrastructure, blog and content infrastructure, performance and Core Web Vitals, mobile and responsive UX, PPC landing pages, analytics and testing infrastructure, and structured data and search presentation.

How do you run work like this when several teams touch the same site?

Rarely is a site this size touched by one team. On this engagement the work ran alongside a second agency, a client-side SEO lead, the client's internal content person, and a developer maintaining sister sites. That is the normal condition for a large account, and it is usually treated as an inconvenience rather than something to design around.

Four practices made it work.

One owner per layer. Template and theme changes had a single owner. Content and local listings had others. The failure mode in multi-party work is not disagreement, it is two parties fixing the same thing in different directions, which produces churn that looks like progress.

A standing internal check-in before the client call. A short internal sync ahead of each client meeting, so the client heard one coordinated position instead of watching their vendors negotiate live.

A weekly client call and a monthly full-team sync. The weekly call handled decisions and blockers. The monthly sync was for the parties who do not need to be in a weekly meeting but do need to be aligned on direction.

A written status digest between meetings, covering what shipped, what is pending, what is blocked and on whom, and what needs a decision. Written down, circulated, and carried forward until resolved. Most of what gets lost in a multi-party account is lost in the gap between meetings.

Structural work specifically rewards this discipline, because template changes have blast radius. A change that improves 2,000 pages can break 2,000 pages, and in a multi-party account the person who made it may not be the person who hears about it.

What results did this produce?

Impressions in Google's generative AI features rose from about 154,000 to about 522,000, comparing the most recent three months against the three months before them. That is roughly 368,000 additional impressions and about 3.4 times the previous rate. The figure comes from Google Search Console's Generative AI features report, which Google still labels beta, in its three-month comparison view as retrieved on 17 September 2026. Both periods are equal length and consecutive, and the report expresses them relative to the retrieval date rather than as fixed calendar ranges.

Organic search users rose 18.4%, from 69,700 to 82,500, comparing January 1 through April 30, 2026 against May 1 through September 16, 2026. The second window is about two weeks longer than the first, and home services demand runs higher in summer than in winter, so some part of that gain is seasonal rather than earned. We state the windows rather than presenting the percentage on its own.

Organic search clicks rose 18% on flat impressions. Search Console records 52,600 clicks over the most recent three months, up 18%, against 7.46 million impressions, down 1%, as retrieved on 17 September 2026. The same volume of search visibility now converts into roughly 19% more visits, which is what structural work looks like in the numbers: the same eligibility, better captured. It is also worth noting where this figure comes from. The 18.4% above is an analytics platform counting users; this is Google's own search console counting clicks. Two systems measuring different things by different methods, landing within half a point of each other.

When the rebuild went live changes how to read all of this. The rebuilt theme launched in mid-July 2026. The most recent three-month window therefore opens about four weeks before the launch, and the period it is measured against sits entirely in front of it. Roughly a quarter of the "after" period predates the change being measured. These figures understate the post-launch effect rather than flattering it.

More than 220 improvements shipped, and the site now runs on a framework where future changes can be made at the template level rather than page by page. That is the part that keeps paying after the engagement ends.

A note on what is not claimed here. Total site traffic over the same span grew far faster than organic search did, and Alan Morgan Group does not attribute that growth to this work, because the organic figures do not support it. The 18.4% is the honest number for the search program. Presenting the larger figure next to a description of SEO work would imply a cause the data does not establish.

What should you expect before you start?

A theme rebuild is a real project, not a tune-up. It touches every template on the site and requires QA across every page type, every breakpoint and every component. Budget for the QA, not just the build.

The risk is concentrated and worth naming. Everything that makes template-level work efficient also makes mistakes efficient. Staging, review and a rollback path are not optional on a site with thousands of URLs.

AI visibility metrics are young. Reporting on impressions in generative results is newer and less standardized than click and impression reporting. Treat the direction and the magnitude as meaningful and the precision as approximate.

The gains are structural, so they persist and they compound. New pages built on the rebuilt framework inherit the improvements rather than needing them applied. That is the argument for doing the unglamorous work first.

Frequently asked questions

What is the difference between SEO and AEO or GEO? Traditional SEO optimizes for placement in a ranked list of links. Answer engine optimization and generative engine optimization target being used and cited inside an AI-generated answer. The underlying work overlaps heavily, but AEO puts far more weight on machine readability: clean markup, correct heading hierarchy, accurate structured data, and consistent patterns across page types.

Can you actually influence whether you appear in AI Overviews? Yes, though not by any single lever. In the engagement described here, impressions in Google's generative AI features went from about 154,000 to about 522,000 across two consecutive three-month periods, about 3.4 times the previous rate, following a full WordPress theme rebuild that corrected markup, structured data, heading structure and page performance across every template on the site.

Why does a WordPress theme affect AI search visibility? The theme generates the underlying HTML for every page built on it. Heading structure, structured data placement, markup cleanliness and asset loading are all theme-level decisions. On a site with thousands of URLs produced by a few dozen templates, a structural flaw in the theme is a structural flaw on every page, and a theme-level fix corrects all of them at once.

How long does it take to see a change? This engagement ran through 2026, with the performance shift visible across two consecutive three-month comparison periods. Structural work does not produce a result in weeks. It produces a step change once the rebuilt framework is live and has been recrawled.

Is this only relevant for large sites? The leverage argument scales with the number of URLs. A 20-page site can be fixed page by page. A site with thousands of location, service and content pages generated from shared templates cannot, which is where template-level work stops being a preference and becomes the only workable approach.

What if another agency is already working on the site? That is the common case and it is manageable with clear ownership per layer, a standing internal check-in before client calls, and a written status digest circulated between meetings. The risk in multi-party work is two parties fixing the same thing in opposite directions, not disagreement itself.

Does this replace regular SEO work? No. Structured data, clean markup and performance are the entry requirements for being usable by an AI answer. Content quality, local search infrastructure and authority still determine whether you are the source worth using once you are readable.

Alan Morgan Group rebuilds the technical foundations under large multi-location and national websites. If your traffic is flat and nobody can tell you whether your pages are even readable by the systems now answering your customers' questions, that is the conversation to have.