SEO & AI Search Spam Tactics
A historical reference to SEO, local search, and AI Search: the tactics, how they work, and their status and boundaries.
| Tactic or myth | How it works | Status and boundaries | Period |
|---|---|---|---|
AI recommendation poisoning / prompt injection | Inserting instructions into “Summarize with AI” buttons or links so that an assistant stores a brand as a trusted source or recommends it. | Manipulation of AI assistant memory, not a Google ranking tactic. Microsoft reported attempts of this kind, with effectiveness varying across platforms and mitigations continuing to change. | 2026 |
“AI-specific technical tricks”: llms.txt, mandatory chunking, or magic schema | Selling particular files, content-chunk formats, or markup as requirements for content to appear in AI answers. | Not requirements for Google AI Search. Google says its generative features do not require llms.txt, content split into small chunks, or special schema. This claim is limited to Google Search; other AI platforms may differ. | 2024–2026 |
Manufacturing “consensus” through apparently independent brand mentions | Spreading fabricated mentions or recommendations across blogs, forums, and other sources to make a brand appear popular. | A weak claim when sold as a shortcut. Google’s guidance says inauthentic mentions do little to help visibility in Google’s AI Search features. | 2024–2026 |
Listicles naming their own publisher “best” to attract AI citations | Publishing category lists, placing the publisher’s own brand at the top, and hoping AI systems use the list when answering recommendation queries. | Can generate temporary citations or mentions, but a citation does not guarantee that the brand will be recommended. An Ahrefs experiment found results varied by brand and over time. | 2024–2026 |
Parasite SEO / site reputation abuse | Placing third-party content on an already strong site primarily so that the content can borrow the host’s ranking signals. | A violation when the purpose is to exploit the host site’s reputation. Freelancer or third-party content is not automatically a violation; the context and purpose of its placement determine the issue. | 2023–2026 |
Expired-domain abuse | Buying an old domain with an established reputation, then filling it with unrelated, low-value commercial content. | Abuse of a domain’s reputation. Using an old domain for a useful new site is allowed; the problem is exploiting its previous reputation for rankings. | 2023–2026 |
Scaled content and low-value programmatic SEO | Producing thousands of pages from templates, minimal data, scraping, automatic translation, or AI without adding meaningful value. | Page volume or AI use is not automatically a violation. The problem is mass production intended to manipulate rankings without providing value for users. | 2022–2026 |
Affiliate product reviews that only summarize specifications | Creating “best” lists without testing, firsthand experience, or original analysis. | A list format or affiliate model is not inherently a problem; the lack of original value is. Helpful reviews should provide analysis or experience beyond copying specifications. | 2020–2023 |
Review markup and FAQ schema added everywhere for rich results | Adding markup to obtain stars or more prominent FAQ displays in search results. | Markup does not guarantee rankings. Google restricted review rich results to certain categories in 2019, then reduced FAQ rich results in 2023. A change in appearance does not automatically mean a ranking penalty. | 2019–2023 |
Local SEO / Google Maps spam | Creating fake or duplicate location profiles, stuffing business names with keywords, buying reviews, or asking only for positive reviews. | Spam and review manipulation. Reviews must come from real experiences; incentives, multiple accounts, and filtering out negative reviews violate Maps policies. | 2010s–present |
E-E-A-T theater | Adding author bios, credentials, or markup and selling them as a direct “E-E-A-T score” for rankings. | A misconception when treated as a score or a single ranking recipe. E-E-A-T is not one specific ranking factor; credibility and experience still need to be evident in the content. | 2010s–present |
Fake freshness | Changing the publication date or the year in a title without meaningfully updating the content. | A cosmetic signal. Google explicitly asks site owners to consider whether they are changing dates simply to make pages appear fresh. | 2010s–present |
Article-length formulas, keyword density, and “LSI keywords” | Adding words or related terms solely to meet a particular content-length or keyword-density formula. | A myth when sold as a magic number. Page length should follow the needs of the topic and its readers, rather than a universal word-count target. | 2010s–present |
Doorway pages and city pages that only swap the location name | Creating many nearly identical pages for different queries or cities, then directing visitors to the same destination. | Doorway abuse when the pages are made primarily to rank and add no unique value for visitors. | 2014–2017 |
Private blog networks (PBNs) and mass guest posting for backlinks | Using networks of sites or guest articles primarily as link vehicles, often with keyword-rich anchor text. | Spam when the primary purpose is link manipulation. Guest posts written for a relevant audience can still serve an editorial purpose. | 2013–2017 |
Exact-match domains as a shortcut | Buying a domain that exactly matches a query and expecting the domain name to lift low-effort content. | A keyword in a domain name is not automatically a problem. Google has a system that reduces the excessive ranking benefit of domains designed to match queries exactly. | Around 2010–2012 |
Content farms, spun articles, and thin pages for every keyword variation | Publishing large volumes of inexpensive articles, rewriting existing content, or splitting similar topics across many pages. | Spam or low-quality content when the pages offer no unique value. Panda was announced in 2011 and later became part of the core ranking systems. | 2008–2012 |
PageRank sculpting with nofollow | Adding nofollow to internal links in the hope that their share of PageRank will be redirected to other pages. | A shortcut based on an incorrect assumption about how PageRank flows. | 2007–2009 |
Link spam: comments, guestbooks, mass directory submissions, link exchanges, paid links, and widget links | Creating large numbers of links or using keyword-rich anchor text to make a site appear popular. | Spam when the main purpose is to manipulate rankings. Google addresses these link patterns through its spam policies and systems such as Penguin. Advertising and sponsored links can be used with the appropriate link attributes. | 2000s–2010s |
Google bombing / link bombing | Many sites use similar anchor text to push a page into search results for a particular query, sometimes as a joke or campaign. | An early example of ranking manipulation through link patterns. In 2007, Google announced changes to reduce the impact of many Googlebombs. | 2000s |
Keyword stuffing and meta-keyword stuffing | Repeating keywords excessively on a page or loading the meta keywords tag with large numbers of terms. | Spam. In 2009, Google said it had already ignored meta keywords for ranking for years. | Late 1990s–early 2000s |
Reading the periods: These are approximate eras of discussion and use, not precise dates when a tactic was invented, worked, or stopped working.
Reading the status: Spam, myths, legitimate techniques used excessively, and changing search features have different implications. A lost benefit does not automatically mean a ranking penalty.