Google’s NotebookLM is being turned into an unlikely parasite SEO host. Public notebook pages are appearing in Google Search for spam-oriented queries covering coupon codes, peptides, adult material, apps and other commercially aggressive topics, allowing spammers to exploit pages hosted on Google’s own infrastructure rather than building authority on domains of their own.
The activity was documented on September 22 by Search Engine Roundtable after examples circulated among SEOs Charles Floate, Gagan Ghotra and Glenn Gabe. The report says Google was indexing more than 12,000 of the public notebook pages at the time of observation. That figure should be treated as an observed search-index estimate rather than an official Google statistic, but the underlying abuse was visible in live search results.
The episode is especially notable because NotebookLM is a Google product. Instead of compromising an unrelated high-authority publisher or placing low-quality pages on an established third-party site, spammers can publish indexable material on a Google-hosted surface and attempt to use that environment to gain visibility in Google Search. Some of the reported examples were also observed in AI Overviews, extending the issue beyond conventional blue links.
Why public NotebookLM pages are attractive to parasite SEO
Parasite SEO broadly describes a tactic in which someone publishes content on a stronger third-party domain or platform in order to benefit from signals and discoverability that would be harder to obtain on a new or weak domain. The technique is not inherently identical to every form of third-party publishing, but it becomes abusive when platforms are flooded with manipulative pages created primarily to capture rankings rather than serve the platform’s intended audience.
Public NotebookLM pages create an appealing environment for that strategy because they are shareable web documents associated with a highly trusted Google-owned service. If those pages are crawlable and indexable, an operator does not need to wait for a new domain to be discovered or establish a conventional publishing site before attempting to appear in search.
Search Engine Roundtable’s examples show the familiar pattern: content far removed from the normal educational and research use cases associated with NotebookLM, targeting queries around discount codes, peptides, pornography, sex products and apps. Gabe noted that users can also inspect the prompt that generated a notebook, making the automation behind some pages unusually visible.
More than 12,000 pages were reportedly indexed
The reported scale is what turns the incident from an amusing SEO loophole into a search-quality problem. Search Engine Roundtable said Google appeared to be indexing more than 12,000 such pages when the report was published. There is no indication in the cited report that Google itself confirmed that number or that every indexed NotebookLM page was spam, so the figure should not be read as an audited count of abusive documents.
What the examples do establish is that public notebooks were indexable and that spam-oriented pages had entered the SERPs. Gabe predicted that the pages would likely disappear relatively quickly, while Schwartz argued that Google should be better prepared for this type of abuse when launching new indexable products. As of the cited report, however, there was no documented official Google announcement describing a cleanup or policy change.
The pattern is familiar. Any platform that combines easy publishing, public URLs and strong domain-level visibility can attract opportunistic SEO activity. The difference here is that the host and the search engine are part of the same company, creating the unusual spectacle of Google Search surfacing spam published through a Google AI product.
The AI Overview angle makes the problem more consequential
Traditional spam ranking in Search is one problem; spam being selected as evidence for an AI-generated answer introduces another. Floate’s original demonstration claimed that the technique had reached an AI Overview using Google’s own hosted content. The reported examples suggest that at least some NotebookLM pages were capable of entering generative-search visibility rather than remaining confined to ordinary results.
This matters because AI Overviews do not simply reproduce the top organic results. Research covered by NetContentSEO found that 29.8% of domains cited in AI Overviews were absent from the co-displayed first-page organic results in a 55,393-query study. Source selection for generative answers therefore represents its own retrieval layer, with opportunities and failure modes that are not perfectly captured by conventional ranking analysis.
If low-quality hosted pages can enter that layer, the impact is not limited to a spam result occupying a blue-link position. The content can potentially become part of the evidence Google uses to synthesize an answer, giving an abusive page visibility in an interface where many users may read the generated response without examining every underlying source.
Google’s own domain does not guarantee trustworthy content
The incident is also a useful reminder about how authority should be interpreted in AI search. A URL hosted by a major technology company is not automatically authored, reviewed or endorsed by that company. User-generated and publicly shared content can exist under highly trusted domains while containing material created entirely by third parties.
That distinction becomes more important as AI systems rely on retrieval pipelines that may encounter public documents, community posts, hosted pages and other user-controlled surfaces. Domain reputation can be a useful signal, but it cannot substitute for page-level assessment of who created the information, why it exists and whether its claims deserve to be used.
For SEO practitioners, the NotebookLM case therefore should not be interpreted as proof that “Google properties rank automatically.” Search systems use many signals, results can change quickly, and spam loopholes are particularly unstable. What it does demonstrate is that a newly indexable platform can temporarily create an attractive surface for large-scale opportunistic publishing.
Parasite SEO becomes more complicated in generative search
In classic parasite SEO, the objective is relatively easy to understand: place a page on an authoritative host, rank for a valuable query and funnel the visitor toward a commercial destination. AI search adds additional outcomes. A third-party page might rank organically, be cited in an AI answer, contribute information without receiving a click or influence which brands and products are mentioned by the model.
NetContentSEO recently covered commercial GEO experiments in which third-party pages accounted for 86% of observed brand mentions. Those experiments were not about spam and should not be conflated with the NotebookLM abuse, but they illustrate why third-party publishing has become strategically important in generative visibility. AI systems can use sources outside a brand’s own domain when constructing recommendations and answers.
That makes platform abuse potentially more attractive: the attacker is no longer competing only for a ranking position but for inclusion in an AI retrieval ecosystem. At the same time, it makes enforcement more important because low-quality content that reaches generative systems can affect users even when they never click the underlying page.
This is a search-quality test for new AI publishing surfaces
The NotebookLM spam episode exposes a broader product-launch problem. When a technology company makes AI-generated or AI-assisted pages public and indexable, it is effectively creating a new publishing platform. Search engines then need to decide how those documents should be crawled, indexed, ranked and potentially used by generative answer systems.
Those decisions cannot be separated from abuse prevention. Easy generation lowers the cost of producing thousands of pages, while a trusted hosting environment can lower some of the barriers to discovery. The combination creates an obvious incentive for spammers to test how far the system can be pushed.
Google may ultimately remove the offending pages, adjust indexing behavior or strengthen quality controls around public notebooks. The September 22 evidence does not establish which response, if any, the company will take. What it does establish is that the loophole became visible enough for spam pages to enter Search and, in reported cases, AI Overviews.
For Google, that creates an unusually circular challenge: its AI publishing product can generate public pages, its search engine can index them, and its generative search layer can potentially surface them again. Spammers have noticed that loop. The next question is how quickly Google closes it.