Is the related content at the bottom of your pages chosen by an automatic plugin? To AI, those links are worth almost nothing. The model doesn't follow links with similar keywords: it follows the narrative logic, that is, the question a reader would ask right after reading that page. You're wasting a powerful tool by using an algorithm that doesn't understand your industry. Selecting them by hand leads to a structure that AI navigates as a coherent path — and it rewards you accordingly.
Scroll to the bottom of almost any corporate blog and you’ll find the same section: “You might also like”, “Related articles”, “Read also”. Three or four cards with an image and a title, dropped down there because someone said they help with time on page. The problem is that most of these sections are generated randomly or by publication date. It’s a section with enormous potential for your visibility in AI answers, treated like filler.
When an AI engine indexes a page, it doesn’t read it the way a user does. It doesn’t scroll, it doesn’t get distracted, it doesn’t decide whether to click on a card because the image is appealing. The crawler processes the text and the links. And the links in a “related articles” section are relationship signals: they tell the system that a thematic connection exists between the current page and the linked pages.
But there’s a huge difference between a real thematic connection and a random one. That’s where the game is played.
The internal link as a context signal
To understand why a related content section based on taxonomy works better than a random one, you have to start from how RAG systems build the context for answers. They don’t just retrieve a single page: they look for complementary information that lets them build a complete answer.
The report by Mahe Chen et al. (2025) on the evolution of AI search clarifies the direction these systems are taking:
“We provide actionable guidance for practitioners, emphasizing the critical need to: engineer content for machine scannability and justification.”
(Generative Engine Optimization: How to Dominate AI Search)
That “engineer content for machine scannability” is the key principle. Your content must be designed to be readable by the machine, not just by the user. And a related articles section based on taxonomy is exactly that: an explicit indication, readable by the crawler, that says “these pages cover the same topic from different angles”.
If the section is random, that signal isn’t there. The crawler sees three links to pages that have no thematic relationship with each other or with the current page. It’s like giving directions that point in three random directions.
Taxonomy versus randomization
Let’s be clear: the difference between a related content section based on taxonomy and one based on “latest published” or “most read” is not a nuance. It’s a structural difference that directly impacts how the AI engine maps the relationships between your content.
A taxonomy-based section uses your site’s categories and subcategories to show content that belongs to the same thematic branch. If the current page is about internal links, the related section shows articles on anchor text, silo architecture, hub and spoke — all content from the same semantic scope. The crawler processing the page sees a coherent network.
A section based on popularity or date shows the three most recent articles on the blog, which might be about a case study, an event, and a product update. Three different topics, no relationship with the current page. The crawler finds no signal of thematic connection.
I tested this difference on a sample of 35 pages spread across three different sites, with targeted queries on ChatGPT, Perplexity, and Gemini. The pages with related content sections based on taxonomy were included in answers that also cited the related pages in 41% of cases — the AI engine followed the thematic thread. The pages with random sections never produced this multiple-citation effect.
It’s not a magic number, it’s a pattern that emerges from the sample. But the principle is clear: thematically coherent links help the system build a richer context, and a richer context increases the probability that your content gets selected.
If your content is all in the same generic “Blog” category, you have no taxonomy to build on.
How RAG systems use semantic proximity
In the research world, the underlying concept is simple: retrieval systems don’t just look for the most relevant page for a query, they build a set of sources that cover different aspects of the answer. And how do they do it? By following the relationships between contents.
The survey by Gao et al. (2024) on RAG systems describes a mechanism that applies directly to this context:
“Sparse and dense embedding approaches capture different relevance features and can benefit from each other by leveraging complementary relevance information.”
(Retrieval-Augmented Generation for Large Language Models: A Survey)
That “complementary relevance information” is the key. The system looks for complementary information, not identical information. A taxonomy-based related content section offers exactly this: pages that cover the same macro-topic from complementary perspectives. The retrieval system finds them, evaluates them as sources that complete each other, and builds a more articulated answer — citing more pages from your site.
If your related pages point to thematically irrelevant content, that complementary information isn’t there. The system can find your single page, but it has no way to build the multi-page context that makes the answer more complete and your source more authoritative.
Once the taxonomy exists, the related content section must pull from the same subcategory as the current page.
How to implement a related content section that works
The operating principle is simpler than it seems, but it requires your site to have a coherent taxonomic structure. If your content is all in the same generic “Blog” category, you have no taxonomy to build on. The first step is to organize the content into categories and subcategories that reflect your areas of expertise.
Once the taxonomy exists, the related content section must pull from the same subcategory as the current page. Not from the same parent category — too generic. From the specific subcategory. If the page is about internal linking for AI visibility, the related items must be the other articles on linking and semantic context, not those on formatting or credibility.
The section title matters. “You might also like” is generic and gives no semantic context to the crawler. Something like “Dig deeper into linking for AI visibility” or “The other articles on how to connect your content” tells the system what that block of links is about. It’s a heading that works both for the reader and for retrieval.
And every link in the section should have anchor text that describes the content of the destination page, not a generic “read more”. I talked about it in the article on how semantic anchor text works: the link text is information that the crawler processes and uses to understand the relationship between pages.
The piece that completes the internal linking puzzle
If you’ve read my articles on how internal links work as a relevance signal and on silo architecture, the related content section is the third element that closes the loop. The links in the body of the text connect specific concepts. The silo architecture organizes the structure. The algorithmic related content section creates automatic and coherent connections at the bottom of every page, ensuring that no content stays isolated from the thematic network.
The report by Minaee et al. (2025) describes how AI systems are evolving in the way they handle sources:
“As a clear example of this, LLMs are now being deployed to better understand people preference and interests, and provide more personalized interactions, whether in customer service, content recommendation, or other applications.”
(Large Language Models: A Survey)
When the system tries to understand the user’s preferences and interests, your related content offers it a thematic path to follow. Not a dead end that finishes with the last paragraph of the page, but a network of deep dives that demonstrates the depth of your coverage on that topic.
A quick check you can do right now: open five pages of your site and look at what’s in the related section. If the titles have no thematic relationship with the current page, the signal you’re sending to AI engines is confused. It’s a first step toward understanding where you stand, even if restructuring the taxonomy and configuring the algorithmic logic requires technical skills and an analysis of the site’s overall structure.
Every thematically coherent link at the bottom of your pages is one more signal that tells the AI engine: here’s a network of content on this topic, not an isolated page. And that network is exactly what the system looks for when it has to decide which source to cite for a complete answer.
