Links: a web of citations
The 1998 Google paper treated the link graph as a web of citations [1]. Its authors wrote that anchors “often provide more accurate descriptions of web pages than the pages themselves” [1]. PageRank, the ranking built on that idea, gives a page weight when other weighty pages are willing to point to it [2]. In the same period, Kleinberg described a related view: authoritative pages on a topic, and “hub” pages that link to many of them [7].
A familiar hope is that more links are simply better. Structures built only to borrow that weight were already described as link spam in 2005. The same study placed the repetition of query words through titles, body text and anchors under the name term spam [3]. Those moves can lift a page for a while. While they do, a careful site can sit lower than it deserves [4]. Google’s current guide adds a plain limit: PageRank is one of its fundamental algorithms, but “there’s much more to Google Search than just links” [14].
There is a more encouraging result beside that warning. In a model of search advertising, optimization improved the unpaid results when a site’s usefulness to people moved with the value the site hoped to earn [5]. Ads then gave weaker sites less reason to dress themselves up, and visitors began with the organic result [5]. In the same model, the search engine itself typically earned less, because money moved from sponsored listings toward unpaid ones [5].
Titles: one clear name for the page
The words on the page still matter, though a density score is a poor way to say how. A 2005 experiment found that visibility rose when the words people search for appeared more often in the title, in the body, or in both [6]. A later reading noted that using them in the title and the body beat optimizing the title alone [4]. We keep the lesson and leave the quota behind. Repeating a phrase past the point a reader would welcome is the stuffing that the taxonomy called spam [3] [6].
Google’s documentation now says how titles are used. Its title links are generated automatically from several sources, including the title element, the main visible title, headings, and the text of links that point to the page [12]. It asks for descriptive, concise titles, warns against keyword stuffing, and notes that when it cannot read a page it may rely on anchor text from other sites [12]. That last point connects back to the 1998 paper [1]. Structured data plays a narrower role. Google describes it as explicit clues about a page’s meaning that can make a page eligible for rich results, and asks that markup describe content visible to the user [13].
Answer passages: matching words and matching meaning
Classic retrieval ranks text by the words it shares with the query. BM25 is the best-known result of the probabilistic relevance framework, and its variant BM25F takes document structure into account [8]. Dense retrieval matches meaning instead. In 2020, a dense passage retriever beat a strong BM25 system by 9 to 19 percentage points in top-20 passage retrieval accuracy on open-domain question answering datasets [9].
That gain did not carry everywhere. Across 18 datasets outside the training domain, BM25 was a robust baseline, and dense models often underperformed other approaches [10]. For a page, the reading is simple. Use the words your reader uses, and say what you mean in one passage. Both kinds of retriever can then find it.
Position inside the passage set also matters once a language model reads it. Liu and colleagues found that performance was often highest when the relevant passage sat at the start or end of the model’s input, and dropped when it sat in the middle [11]. A site doesn’t choose where an engine puts its text. It can make sure the answer is not buried, so the passage that gets retrieved is the one that answers.
How these habits compare
| Habit | What the research reported |
|---|---|
| Links built only to borrow weight | Described as link spam in 2005 [3]. |
| A link whose text describes the page | Treated in 1998 as a description of the target page [1]. Still one of the sources Google uses for title links [12]. |
| Repeating the query through the page | Described as term spam in the same taxonomy [3]. Against Google’s spam policies [14]. |
| Search words in the title and the body | Visibility rose in a 2005 experiment [6]. Not a density quota for the engines people use now. |
| One passage that answers the question | Found by word-matching and meaning-matching retrievers [9] [10]. Used best by models when it isn’t buried mid-input [11]. |
Ranking systems move, and a finding from 1998 or 2005 is not a recipe for this year [1] [6]. When a tactic has to be hidden from the person reading, it is usually on its way to being hidden from the engine as well [3] [4].
Map
Look at the map. For each query cluster, note who holds the results and what their titles promise. Note which pages others link to, and with what words [1] [12].
Find your place. Check whether your title says what the page is for. Check whether one passage answers the question in the reader’s words, and whether anyone who has read you links to you [6] [10] [12].
Align. Rename unclear titles. Move the answer to the top of its section [11]. Earn links by being worth citing, not by building them [3] [4]. Mark up only what the reader can see [13].
References
- Brin, S., & Page, L. (1998). The anatomy of a large-scale hypertextual Web search engine. Computer Networks and ISDN Systems, 30(1–7), 107–117. https://doi.org/10.1016/S0169-7552(98)00110-X
- Page, L., Brin, S., Motwani, R., & Winograd, T. (1999). The PageRank citation ranking: Bringing order to the web (Technical Report 1999-66). Stanford InfoLab. http://ilpubs.stanford.edu:8090/422/
- Gyöngyi, Z., & Garcia-Molina, H. (2005). Web spam taxonomy. Proceedings of the First International Workshop on Adversarial Information Retrieval on the Web (AIRWeb). https://airweb.cse.lehigh.edu/2005/gyongyi.pdf
- Malaga, R. A. (2008). Worst practices in search engine optimization. Communications of the ACM, 51(12), 147–150. https://doi.org/10.1145/1409360.1409388
- Berman, R., & Katona, Z. (2013). The role of search engine optimization in search marketing. Marketing Science, 32(4), 644–651. https://doi.org/10.1287/mksc.2013.0783
- Zhang, J., & Dimitroff, A. (2005). The impact of webpage content characteristics on webpage visibility in search engine results (Part I). Information Processing & Management, 41(3), 665–690. https://doi.org/10.1016/j.ipm.2003.12.001
- Kleinberg, J. M. (1999). Authoritative sources in a hyperlinked environment. Journal of the ACM, 46(5), 604–632. https://doi.org/10.1145/324133.324140
- Robertson, S., & Zaragoza, H. (2009). The probabilistic relevance framework: BM25 and beyond. Foundations and Trends in Information Retrieval, 3(4), 333–389. https://doi.org/10.1561/1500000019
- Karpukhin, V., Oğuz, B., Min, S., Lewis, P., Wu, L., Edunov, S., Chen, D., & Yih, W. (2020). Dense passage retrieval for open-domain question answering. Proceedings of EMNLP 2020. https://aclanthology.org/2020.emnlp-main.550/
- Thakur, N., Reimers, N., Rücklé, A., Srivastava, A., & Gurevych, I. (2021). BEIR: A heterogeneous benchmark for zero-shot evaluation of information retrieval models. NeurIPS 2021 Datasets and Benchmarks Track. https://arxiv.org/abs/2104.08663
- Liu, N. F., Lin, K., Hewitt, J., Paranjape, A., Bevilacqua, M., Petroni, F., & Liang, P. (2024). Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 12. https://aclanthology.org/2024.tacl-1.9/
- Google Search Central. (n.d.). Influencing your title links in search results. Google for Developers. Accessed 9 October 2026. https://developers.google.com/search/docs/appearance/title-link
- Google Search Central. (n.d.). Introduction to structured data markup in Google Search. Google for Developers. Accessed 9 October 2026. https://developers.google.com/search/docs/appearance/structured-data/intro-structured-data
- Google Search Central. (n.d.). Search engine optimization (SEO) starter guide. Google for Developers. Accessed 9 October 2026. https://developers.google.com/search/docs/fundamentals/seo-starter-guide
When someone searches for what you do, which title, passage and link do you want them to find? Tell us where you want to stand. Or continue with what a generated answer notices.