A survey covering 45 studies finds that generative engine optimization (GEO) rewrites can reduce a page's AI retrieval by 16%. PPC Land reported the figure. It is one of the first aggregated numbers to quantify what many publishers have suspected anecdotally: when a generative engine rewrites and repackages a page's content, the original source can lose visibility inside AI answers.

The number matters because it converts a vague industry anxiety into a measurable effect. Until now, most GEO commentary relied on single-tool screenshots and isolated rank checks. This survey pools 45 separate studies into one meta-level finding. That scale separates it from the tool-signal speculation that dominates the topic.

What does the 16% figure actually measure?

The finding concerns AI retrieval — whether a generative engine pulls in, cites or otherwise surfaces a specific page when it constructs an answer. A 16% reduction means that, across the studies surveyed, pages subject to GEO-style rewrites saw their retrieval drop by roughly one-sixth on average.

The mechanism is structural, not algorithmic punishment. Generative engines synthesize content into their own answer text. The more completely an engine rewrites a page's information, the less it needs to retrieve or attribute that page again. The survey's framing — rewrites can cut retrieval — signals a probabilistic effect, not a guaranteed outcome for every URL.

Readers should note what this finding is and is not. It is an aggregation of published research, reported by PPC Land. It is not a confirmed statement from Google, OpenAI, Perplexity or any engine vendor. No search engine has acknowledged that rewriting reduces source retrieval, and none has published a counter-figure.

How strong is the evidence base?

The survey's core strength is breadth: 45 studies, not one dataset. Pooling studies dampens the noise that plagues single-crawler experiments, where one tool's index freshness or one vertical's volatility can skew results.

Its core limitation is aggregation itself. PPC Land's report gives a headline figure — 16% — but the underlying studies likely vary in method, engine tested and content type. Averaging across 45 heterogeneous studies produces a central tendency, not a prediction for any single site. Publishers in verticals with heavy informational intent, such as health, finance and how-to content, face more rewrite exposure than transactional or brand-specific pages, though the report does not break the figure down by vertical.

The confirmed fact here: a 45-study survey found GEO rewrites can cut AI retrieval 16%. The interpretation — how that loss distributes across engines, niches and page types — remains open, and PPC Land's report does not settle it.

What should publishers and SEO teams do with this?

The finding gives teams a baseline for measurement. If a page loses roughly a sixth of its AI retrieval after rewrite-style treatment, that number becomes the benchmark against which mitigation tactics can be tested.

Practical monitoring follows directly from the data:

• Track AI retrieval per URL, not just classic rankings, since the 16% effect concerns generative answers rather than the ten blue links.
• Compare retrieval before and after major content rewrites by engines, using the 16% figure as an expected-magnitude reference.
• Separate confirmed engine announcements from third-party survey findings when reporting internally — this 16% figure belongs to the second category.

For editorial teams, the number quantifies the stakes of answer-engine exposure. A page can rank first in classic results and still lose retrieval share inside a synthesized answer. The survey suggests the two surfaces can move independently, and that rewrite intensity is a variable worth tracking on its own.

What comes next?

Watch for two things: replication studies that test whether the 16% retrieval drop holds for specific engines and verticals, and any vendor response — from Google's AI Overviews team or competing answer engines — addressing how they attribute and retrieve rewritten sources. PPC Land's survey sets the number; the next round of studies will test whether it holds.