A site can grow from 100 articles to 300 and then 1,000 without actually becoming a better media product.
In the worst case, only the warehouse got bigger.
Good articles sit in boxes at the back, the entrance keeps showing the same popular items, and anything not explicitly searched for effectively disappears.
A better mental model is simple:
a website is a store, not a storage room.
The home page is the storefront, topics are aisles, search is the clerk, article cards are shelf labels and related articles point to the next shelf. Watch where customers go, then change the arrangement if good products are being missed.
The ideal store does not trap customers far from the exit.
It lets someone find one thing quickly, browse deeply when they feel like it, return because something remains worth exploring, and gradually improves the shelves by learning what people could and could not find.
That is the loop:
find → read → discover a next interest → return → improve automatically.
1. Did the reader find what they needed, not just stay longer?
Page views and dwell time are dangerous when treated as goals by themselves.
Someone who gets a one-sentence answer in 30 seconds and leaves may have had an excellent experience.
Someone who wanders across five pages because the answer is hidden may produce five page views while having a terrible one.
The HEART approach cited in the original article separates satisfaction, use, new users, return visits and whether people accomplish what they came to do. It starts with what the product should help people achieve, then chooses signs and numbers that show whether that happened.[1]
So before celebrating “more pages per session,” separate:
- task completion,
- interesting discovery,
- finding a useful next step,
- willingness to return.
The thing worth maximizing is useful discoveries.
If page views rise as a result, great. If page views become the drug, you eventually design a shopping mall with no exits.
2. Find: people follow information scent, not raw link counts
The information-search research cited in the original article describes people using clues to decide where the information they want is likely to be.[2]
On the web, those cues create information scent.
A card saying only “Work,” “AI,” or “Life” tells the category but not the payoff.
A card saying “How to refuse a request without wrecking the relationship” tells the user what may happen after the click.
Discovery quality therefore depends on whether the title, description, topic, user situation, and expected value all point in the same direction.
The Google guidance cited in the original article recommends crawlable internal links with concise, relevant anchor text, and says there is no magical ideal number of links on a page.[3]
Wanting every article to be discoverable does not mean placing every product at the checkout counter.
3. Read: do not force every article into the same template
“Give the answer in five seconds” is useful advice for some content and destructive advice for other content.
Troubleshooting, procedures, pricing, and setup guides often benefit from an early answer.
Essays, observations, cultural pieces, travel stories, and narrative analysis may derive value from the path itself.
Define the reader's job by content type:
- Finding an answer: give the answer, conditions and steps early.
- Choosing: explain the differences and what is good or difficult about each option.
- Understanding why: walk through why something happens.
- Reading a story: keep the events, surprises and discoveries.
- Avoiding serious mistakes: show the evidence, when the advice applies and what remains uncertain.
Google's people-first guidance cited in the original article likewise emphasizes content made for an actual audience, useful enough to help people achieve their goals and leave satisfied, rather than content manufactured primarily to manipulate rankings.[4]
Choose the format from the human goal, not from an SEO superstition.
4. Next interest: answer first, then reveal one useful unknown
A good list of related articles does more than ask for one more page view.
Its job is to make visible a question the reader had not yet articulated.
Loewenstein's information-gap theory proposes that curiosity can arise when people become aware of a gap between what they know and what they want to know.[5]
The design implication is not to hide the current answer.
Answer the current question.
Then offer three directions:
- go deeper,
- prepare for the next likely problem,
- explore an adjacent surprise.
This is the opposite of clickbait.
Not “the answer is on the next page,” but “the answer is here; now that you know it, this is the next interesting problem.”
The cited recommendation research warns that showing already-popular articles, then using those reactions to choose what to show next, can give those articles an even bigger advantage and leave readers seeing less variety.[6]
Alongside articles that fit the topic, give some space to:
- new content,
- good articles that few readers have seen,
- adjacent topics,
- different articles, so the same few do not stay at the front forever.
An article nobody has seen may not be unpopular. It may simply never have had a chance.
5. Return: make continuation valuable, not compulsory
You do not have to start with notifications or registration popups just because you want readers to return.
The self-determination research cited in the original article treats choosing for yourself, feeling capable and connecting with others as important to motivation that comes from within.[7] That does not prove these changes will increase return visits, but it offers ideas to try.
- Let people choose, change or remove their interests without forcing registration.
- Help them finish a visit feeling that they understand, can choose or can do something they could not before.
- Make it clear who writes the site and why, with a consistent approach across articles.
Useful ways back include saved articles, recent reading, an unfinished article, new pieces in chosen topics and something that naturally follows a previous read.
The ideal is not “come back or miss out.” It is come back and take the next step from where you reached last time.
6. Fewer choices are not automatically better
UX folklore often turns “too many choices can be difficult” into a universal law.
The evidence is messier.
Scheibehenne and colleagues combined results from 50 studies, 63 conditions and 5,036 participants. On average, extra choice had almost no effect on difficulty choosing, while results varied considerably between studies.[8]
So research does not give us a universal command such as:
“always show three related articles,” “never have more than seven categories,” or “eight options melt the brain.”
The real questions are whether:
- the differences make sense,
- choices are grouped,
- the user has a meaningful entry point,
- readers know what to compare,
- choosing now is actually necessary.
One hundred articles in a flat list can be awful.
One hundred articles organized around “I have a work problem,” “I want something interesting,” “I want to use AI,” and “I want to make life easier” can become an asset.
7. Automatic improvement: training only on clicks makes the store monotonous
Before improving the site automatically, agree on what each count means. Adding AI comes later.
For each article, record what was shown and where, whether it was clicked, whether the reader continued afterward, whether they saved it or marked it useful, and whether they returned. Also track articles that were never shown.
Keep the article ID, language, text version and recommendation version together. Mix responses to yesterday’s text with responses to today’s rewrite and you will not know what improved.
Publish → show → watch the response → collect results → consider why → change a little → check again.
But do not chase clicks alone. Sensational titles, extreme topics and already-popular articles can win that contest.
Also ask whether readers achieved their purpose, liked the recommendation and returned. Check that buried articles get a chance, topics remain varied and pages stay fast and readable.
Automation should not stop at “more clicks, we win.” Keep checking that the shop is not showing the same shelf forever.
8. SEO and UX are not enemies, unless Google becomes the only customer
The Google guidance cited in the original article emphasizes useful original content, words readers actually use and links search engines can follow.[4][3]
Fast, easy-to-use pages matter, but speed alone does not determine search rankings.[9] The web.dev guidance cited there gave these good thresholds at the 75th percentile: at least 75% of visits should fall within them, rather than judging only a few fast visits.
- LCP: main content appears within 2.5 seconds.
- INP: the page responds to an interaction within 200 milliseconds.
- CLS: the measure of unexpected layout movement is at most 0.1.[9]
For multilingual sites, make clear which pages are different language or regional versions of one another. The cited guidance describes hreflang as a way to specify those connections.[10]
The AI-search guidance cited in the original article also favored useful original material over generic recycling.[11]
Good titles, meaningful links between articles, quick display, sources, clear updates, suitable language URLs and original experiences or analysis help readers as well as search engines.
9. The store metaphor clarifies the whole system
Thinking of the site as a shop makes each job easier to understand.
- Home page: storefront, showing what is interesting now.
- Topics: aisles, grouping articles for similar purposes.
- Search: clerk, finding a destination from the reader’s words.
- Article: product, worth reading on its own.
- Related articles: next shelf, offering depth, the next problem or a detour.
- Saves and history: items put aside and shopping notes.
- Popularity ranking: sales records, but being shown is not the same as being chosen.
- Visit records: seeing where customers went.
- Article creation: back room.
- Checking articles for mistakes: inspecting the goods.
- Publishing: putting goods on the shelves.
- Updating recommendations: rearranging those shelves.
- Helping searchers find the site: signs and maps.
- Other languages: walking the same shop in another language, rather than building a different shop.
Then the owner does not have to move every article by hand each day. Decide what kind of shop it should be, and check that the automatic shelf changes are not doing something strange.
10. The end state is not a site that traps people. It is a site that learns
The desired loop looks like this:
- Value is understandable from search, topics, referrals, or recommendations.
- A single article can fully solve the current question.
- After the answer, a natural next question becomes visible.
- Only people who want to continue are invited onward.
- Returning readers can quickly resume.
- Record what was shown, clicked, read next, revisited or never shown.
- Adjust suggestions and shelves without always favoring already-popular articles.
- After each change, check the writing, speed, search, language versions and whether the articles were published correctly.
- Changes that fail can be reversed.
- The system improves not just articles, but the probability that a good article will be found.
Then article count stops being inventory for inventory's sake.
More articles create more meaningful next steps. More readers create more evidence about navigation. More evidence reveals buried content. Better discovery reduces dependence on search alone.
A strong site does not imprison readers. It gets them to what they need quickly, helps them encounter something they did not know they wanted, and is slightly smarter when they return.
Stop merely expanding the warehouse.
Build the store whose shelves learn.
References (11)
- Measuring how people use a site research.google
- How people search for information doi.org
- Links people and search engines can follow developers.google.com
- Helpful content developers.google.com
- Curiosity cmu.edu
- Recommendations and popularity bias arxiv.org
- Choosing for yourself and motivation selfdeterminationtheory.org
- Number of choices and difficulty choosing scheibehenne.com
- Page experience and loading speed: ; https://web.dev/articles/vitals developers.google.com
- Pages in different languages developers.google.com
- Original article’s reference on AI search developers.google.com
