Understanding Generative Engine Optimization: What the Data Actually Shows

Generative Engine Optimization (GEO) is the practice of structuring content so AI systems — ChatGPT, Perplexity, Gemini, Google’s AI Overviews — can read, trust, and cite it directly in an answer instead of just linking to it. That’s a real shift in how content gets found. But GEO sits on top of the same infrastructure traditional SEO depends on: crawling, indexing, and a canonical URL Google is willing to store. In a September 2026 review of Google Search Console URL Inspection data across seven websites we operate, 28% of inspected URLs (419 of 1,482) were not in Google’s index at all. No AI system that leans on Google’s index can cite a page Google never stored, no matter how well the copy is written for GEO.

The Indexing Problem GEO Content Doesn’t Fix

Of those 419 not-indexed URLs, 185 were pages Google had discovered but not crawled, 57 had been crawled but not indexed, and 120 were unknown to Google entirely. None of those pages can appear in an AI Overview or be cited by an outside model that relies on Google’s index — the quality of the writing on the page doesn’t matter if Google never stored it.

On one site, the sitemap hadn’t been downloaded by Google in eight weeks; in that window the site grew from 209 to 279 URLs, and 84 of them were unknown to Google. On the same property, 79 sitemap URLs reported a last-modified timestamp equal to the exact moment each request was made. Google states it uses lastmod only “if it’s consistently and verifiably… accurate”. Separately, on allcleardigital.com, for 11 products published at two different URLs, Google selected the thinner store stub as the canonical version over the fuller product page — in one measured pair, 492 words against 81.

Indexing isn’t permanent, either. One article in the same audit had earned 2,063 Search Console impressions before dropping to Discovered – currently not indexed. Whatever authority it had built up didn’t protect it once Google re-evaluated it.

How Much Traffic AI Assistants Actually Send Right Now

Even where GEO succeeds, it’s worth sizing the payoff before rebuilding a content strategy around it. Across seven sites over a 90-day window, sessions attributed to ChatGPT, Perplexity, Gemini, Claude, or Copilot accounted for 26 of 3,842 total sessions — under 1%. That’s not an argument against GEO. It’s an argument against treating it as the primary optimization target ahead of basic indexing health, which is still where most of the visible traffic problem lives.

What Google Actually Says About Scaled AI Content

Google’s spam policy defines scaled content abuse as “many pages generated for the primary purpose of manipulating search rankings and not helping users,” and specifically names “using generative AI tools or other similar tools to generate many pages without adding value for users” and “automated transformations like synonymizing, translating, or other obfuscation techniques” as examples. It’s an enforced category, worth reading before scaling up output.

There’s outside data pointing the same direction. Lily Ray tracked 220+ sites named as customers of AI content platforms — third-party estimates from Ahrefs and Sistrix, correlation rather than proven causation — and found 54% lost at least 30% of their peak organic traffic, 39% lost at least half, and 22% lost at least 75%, typically within a year of the peak. That pattern is worth weighing before treating AI-generated volume as a growth strategy on its own.

One Misconception Worth Correcting

Submitting URLs through Google’s Indexing API doesn’t speed up indexing for ordinary blog or service pages. Google states that “the Indexing API can only be used to crawl pages with either JobPosting or BroadcastEvent embedded in a VideoObject”. If a GEO checklist includes pinging the Indexing API for regular content, it isn’t doing what it appears to be doing.

Where GEO Actually Fits

None of this means GEO is pointless — it means indexing health has to come first, and GEO techniques apply on top of a page Google can already find. Among the named services on this site, Agent-Readiness Optimization is described as work to “make your site easier for AI agents, browser assistants, and answer engines to understand and use,” AI Search Monitoring & Citation Recovery is described as work to “track where your brand appears in AI answers and build the pages needed to win lost citations,” and GEO & AI-Citable Copywriting is described as “high-density marketing copy structured with Princeton-validated GEO signals.” Any of that work is only as useful as the indexing status of the page it’s built on.