usdigitalmarketingindex.com SEO
How can automatic related-post widgets create contextually irrelevant internal links?
Updated November 15, 2026
By choosing posts with a rule that measures something other than what the reader needs next. A related-posts widget picks a handful of links for the end of every article using shared tags, shared categories, word overlap, recency, or popularity. Each of those can pair two posts that have a label in common and nothing useful to say to the same reader. A post about fixing a leaking tap is matched with company news because both carry the tag plumbing. An article for beginners is matched with one for specialists because they share vocabulary. The widget then places those links on every page automatically, so one weak rule produces weak links at the scale of the whole site.
Topic similarity versus user task
Two posts can be similar in topic and unrelated in task.
| Selection rule | What it measures | How it goes wrong |
|---|---|---|
| Same category | Filing | Categories are broad; everything in "Blog" is related to everything else |
| Shared tags | Labels | Tags are applied loosely and inconsistently |
| Text similarity | Shared words | Matches near-duplicates and posts that reuse boilerplate |
| Most recent | Date | Has nothing to do with the article |
| Most popular | Traffic | The same few posts appear everywhere |
| Random within category | Nothing | Arbitrary |
The reader’s question at the end of an article is: what do I need next? The answer is a post on the next step, a prerequisite, an alternative, or a deeper treatment of something the article mentioned. Similarity rules do not ask that question.
Failure patterns in auto recommendations
Patterns to look for on your own site.
- Label matches with no task relationship: a how-to matched with a press release.
- Wrong audience: a consumer guide matched with trade content.
- Wrong stage: a post for someone choosing a product matched with one for someone troubleshooting it.
- Outdated targets: links to posts about discontinued products, past events, or superseded advice.
- Near-duplicates: the widget recommends an older version of the same article.
- Popularity loops: a few popular posts are linked from everywhere, gain more views, and stay popular.
- Geographic mismatch: an article about one region matched with rules for another.
- Self-reference and repetition: the same three links on every post in a category.
- Promotional intrusion: the slots are filled with product pages unrelated to the article.
The cost is more than wasted space. Internal links describe how a site’s pages relate. Google’s link guidance asks site owners to “Think about what other resources on your site could help your readers understand a given page on your site, and link to those pages in context” (Google’s guidance on crawlable links and anchor text). A widget that links at random describes relationships that do not exist, on every page.
Editorial rules
Improve the selection logic in layers, from the cheapest change to the most deliberate.
- Exclude. Remove from the candidate pool: news and announcements, expired or outdated posts, noindexed pages, posts flagged as superseded.
- Narrow the pool. Draw candidates from the article’s own topic cluster or subcategory, not from the whole blog.
- Clean the taxonomy. Merge duplicate tags and retire ones attached to a single post. A widget is only as good as the labels it reads.
- Add rules by content type. A how-to may recommend how-tos and reference pages, not company news.
- Allow manual override. Give editors a field to choose the related posts for an article. Use the automatic rule only where the field is empty.
- Set a minimum. If fewer than two relevant candidates exist, show fewer links or none. An empty slot is better than a wrong link.
- Cap repetition. Limit how many times one post can be recommended across the site.
For important articles, choose by hand. The manual override is where the next-step question gets a real answer.
Decide also whether the widget is needed. Links placed in the body, at the sentence where a related subject comes up, carry their context with them. A widget at the foot of the page is a supplement to those, not a replacement.
Validation samples
Test the widget by reading its output. A sample for one source article, with each recommendation judged. The posts are illustrative.
Source article: How to fix a dripping kitchen tap.
| Recommended post | Relevant? | Reason |
|---|---|---|
| How to replace a tap washer | Yes | A deeper treatment of the step the article describes |
| When to replace a tap instead of repairing it | Yes | The alternative, for readers whose repair fails |
| How to turn off your water supply | Yes | A prerequisite |
| We have opened a new showroom | No | Company news; shares the tag "plumbing" |
| Commercial kitchen plumbing regulations | No | Different audience; shares the word "kitchen" |
| How to fix a dripping tap (2019) | No | An older version of the same article |
| Top 10 bathroom design trends | No | Popular post, unrelated task |
Three relevant and four irrelevant, out of seven recommendations.
To run the test across a site:
- Pick a sample of source articles: several from each category, including old and new.
- Record the links the widget shows for each.
- Judge each link with one question: would a reader who just finished the source article be helped by this?
- Give the reason, as in the table, so that patterns show.
- Calculate the share judged relevant, and group the failures by cause.
- Change the rule that causes the largest group of failures, and repeat on the same sample.
Check the mechanics as well.
| Check | Pass condition |
|---|---|
| Links in rendered HTML | Anchors with href, present without interaction |
| Destinations | Return 200, no redirects |
| Excluded content | Never appears |
| Variety | No single post dominates |
| Consistency | Reloading does not shuffle links at random, unless rotation is intended |
Use analytics as a second opinion. Click-through on the widget, by source article, shows which pairings readers found useful. A widget that almost nobody clicks is a candidate for redesign or removal.
Automatic widgets guess at relationships between articles. Deciding deliberately when two articles in the same topic cluster should link to each other is a matter of editorial judgment, set out in the guide to when two supporting articles in one topic cluster should link to each other.
The test: read what the widget recommends as if you had just finished the article. If you would not click any of them, neither will anyone else.
Add your comment to this answer