Thanks for reading. Feedback?
Reply to this email or message [email protected]

Let’s get into it

Before You Publish Another Article

Audit what you already own. Some low-traffic pages should disappear, while others may be earning
valuable AI citations.

Most content teams are built to add things.

New product feature? Publish.

New AI-search question appearing in your prompt research? Create another page.

Much less organizational machinery exists for making the opposite decision: this page no longer deserves to be here.

QuickBooks made that decision at unusual scale.

The company removed more than 2,000 posts from its Resource Center, eliminating over 40% of the library. Within a few weeks, organic traffic was up 20%. During the later tax-season peak, traffic was up 44%, with the additional traffic contributing to a reported 72% increase in signups.

Don’t stop reading here and use this as a takeaway message as “delete half your blog.” Trust me, you’ll want to keep reading.

QuickBooks was cleaning up a Resource Center assembled from several earlier blogs, and some of the later traffic growth coincided with tax season. This wasn’t a controlled experiment proving that deleting URLs causes traffic to flow neatly into whatever remains.

Now for some interesting backstory: the experiment was done in 2018. That feels like a decade ago, because well, it pretty much was.

The experiment is interesting for a different reason now.

In 2018, producing another 100 articles required a meaningful amount of money, time, and editorial capacity.

In 2026, producing another 100 articles can be frighteningly easy.

The constraint has changed.

Before adding another article to the editorial calendar, content teams increasingly need to ask a different question: are we missing an important buyer question, or are we about to publish a fifth version of an answer that already exists somewhere on the site?

QuickBooks didn’t just delete unpopular articles

The experiment makes the project sound more reckless than it was.

QuickBooks had inherited old posts, broken links, overlapping articles and lightweight coverage from several earlier blogs while its team continued adding newer content.

Newer articles that hadn’t had enough time to establish themselves were protected.

The team then investigated posts with fewer than 100 visits during the previous six months and no conversions. That narrowed the library to roughly 2,000 candidates. Pages with useful backlinks or obvious successor content could be redirected rather than simply erased.

The principle wasn’t “low traffic equals bad content.” It was closer to this:

Every URL needs a job.

Traffic can be that job. So can conversion, backlinks, sales enablement, buyer education, strategic positioning, or AI-search visibility.

What becomes harder to defend is a URL whose only remaining purpose is proving that the company once published it.

Userpilot made the same decision with content that was still getting traffic

This isn’t only about an old QuickBooks story.

Userpilot VP of Marketing Emilia Korczynska reported that the B2B SaaS company removed or merged 847 blog posts, approximately 23% of its library.

Those articles were not invisible…Collectively, they had generated more than 24,000 visits during the first eight months of 2024.

Userpilot removed them anyway.

After the pruning project, Korczynska reported a 16% increase in traffic and said the company was on pace at the time for its highest-ever visitor count.

The interesting number here isn’t 847…It is one.

Across the pages Userpilot identified for pruning, Korczynska said the group had produced one conversion in a year.

Among them were hundreds of articles receiving fewer than 10 monthly visits, pages without meaningful backlinks, zero-volume keyword combinations left over from programmatic SEO experiments, topics whose search intent had converged, and content that no longer fit Userpilot's move upmarket.

That last category should make you uncomfortable.

Your content library can remain factually accurate while becoming strategically wrong.

Congratulations. You may now have an extremely efficient acquisition engine for people the business no longer cares much about acquiring.

AI changed the economics of keeping everything

For most of modern SEO, content expansion made intuitive sense.

Keyword research exposed an opportunity, a brief became an article and the article gave the site another potential search entry point.

Generative AI has made the production side of that system dramatically easier.

Google hasn’t said AI-generated content is inherently bad. Its policies focus on whether large quantities of content are being created without meaningful value or primarily to manipulate search visibility.

The AEO implication is more important.

Google has specifically warned publishers against producing separate pages for every possible variation of a question simply because generative search may fan out across related queries. A larger collection of narrowly differentiated URLs doesn’t automatically make a site more relevant.

That should challenge one of the easiest AEO strategies to sell internally.

You find:

“What is X?”

Then:

“X vs. Y.”

Then:

“Best X for enterprise.”

Then:

“Best X for healthcare.”

Then somebody realizes AI can generate another 600 combinations before lunch. The technical ability to manufacture those pages doesn’t mean your website needs them.

The unit of strategy should increasingly be the buyer question and the quality of your answer, not the number of keyword variations you can convert into URLs.

The difficult pages are not the ones with zero traffic

Deleting a six-year-old event recap with no traffic, links, or strategic purpose isn’t a sophisticated content decision. That should be an easy one to make.

The hard decisions start when a page is doing something, just not necessarily something useful.

Consider two articles.

One receives 60 organic visits a month. It answers a highly specific implementation question that routinely appears late in enterprise evaluations. Sales reps send it to prospects, and AI assistants sometimes cite it when buyers investigate deployment requirements.

Another receives 12,000 visits a month. Most visitors arrive through a broad informational query that attracts people outside your ICP. The article generates almost no pipeline and supports a market the company is no longer prioritizing.

Which is the stronger asset? That depends on the type of audit you do.

A traditional content audit can make the second page look untouchable.

A strategic content audit should at least make you ask harder questions.

This is why pageviews alone are a poor indicator of death.

Give every page one of four outcomes

A pruning audit should force a decision about what happens next to each page, not just whether it survives. If your audit ends with only KEEP and DELETE, it’s probably flattening several very different content problems into the same decision.

Most mature libraries require four decisions.

Protect

Protect a page when it performs a distinct job you would notice if it disappeared.

That may include meaningful organic demand, conversions, valuable backlinks, sales usage, ownership of an important buyer question, strategic category coverage or repeated citation in AI answers your buyers encounter.

The traffic number matters, but the function matters more.

Consolidate

Some of the weakest pages in a library are perfectly competent articles that should never have existed separately.

Perhaps your company has a security guide, security FAQ, enterprise security article, and implementation-security page created by four teams over three years.

Each contains part of the answer, but none has become the answer.

AI search makes that fragmentation even harder to justify.

For commercially important subjects, someone inside the team should be able to answer quickly:

Which page contains the version of this idea we want buyers, search engines and AI systems to encounter?

If the answer is “it depends which of our seven articles they find,” you probably have a consolidation opportunity.

Refresh

Poor performance doesn’t always mean the topic lost value…Sometimes the answer deteriorated.

Imagine an integration page that still ranks for a valuable implementation question but describes the product as it worked in 2023.

Deleting it throws away established demand. The trick is to recognize that keeping it untouched gives buyers stale product information.

The right move is to preserve the question and replace the answer.

This difference matters because “obsolete content” can describe two completely different problems:

The subject is obsolete.

Or your treatment of the subject is obsolete.

Analytics can make those pages look remarkably similar.

Retire

Then there are pages where the case for keeping them collapses. These pages look like:

• Programmatic pages built around a keyword combination with no meaningful demand

• Obsolete product announcements

• Thin definition articles completely superseded by a stronger guide.

• Posts attracting the wrong audience for a strategy the company abandoned

• Machine-generated articles somebody approved during a publishing sprint, and nobody has thought about since

The money used to create those assets is irrelevant to the next decision.

The question is whether the organization should continue allocating crawl paths, internal links, update cycles, analytics attention, and editorial judgment to them.

Sometimes the answer should simply be no.

AEO makes the audit harder

Traditional pruning frameworks were built primarily around search traffic…That’s no longer enough.

A technically narrow page can look weak in Google Analytics while playing an outsized role in AI-assisted research.

Suppose an implementation article receives only a few dozen visits a month but repeatedly surfaces when prospects ask an AI assistant whether your product can satisfy a specific technical requirement.

Removing it because it failed your traffic threshold would be an AEO mistake disguised as SEO housekeeping.

For important topics, add AI visibility to the audit before making a decision.

Run representative buyer prompts across the AI systems your audience actually uses. Repeat them enough times to avoid treating one answer as a permanent truth.

Then record more than “cited/not cited.”

Look at which page is being used, which question triggers it, and what happens after consolidation or removal.

If you retire one page, does another page from your domain replace it?

Does your brand disappear from the answer?

Does a competitor or third-party source take its place?

Those are much more useful observations than inventing a metric such as “LLM crawl budget” or assuming authority automatically redistributes when URLs disappear.

We don’t have strong evidence for that mechanism…Prune based on things you can observe.

Don’t sell this internally as a deletion project

Content teams have another problem that QuickBooks' traffic chart doesn’t show.

Publishing is visible work. Deletion isn’t.

Saying, “we published 180 articles” into a quarterly presentation. Or, “we discovered that 63 articles should never have existed independently and merged them into 14 better resources” requires considerably more explanation.

There is also history attached to those URLs.

They are often tied to past decisions, budgets, and internal wins, which means removing them can feel less like optimization and more like questioning the judgment of the people who created them.

So change the conversation.

The objective is to decide where the next dollar and hour of content investment belong.

That is still content production, and it stops defining production as “new URLs created.”

Run a controlled pruning sprint

You don’t need to be as drastic as QuickBooks and remove 40% of the library.

Pick one part of the site where the evidence is unusually clear.

Before changing anything, capture:

• Organic traffic, impressions, and conversions.

• Relevant backlinks and internal links.

• The important buyer questions the content is supposed to answer.

• Current AI-search citations or mentions for representative prompts.

• Overlapping pages that could absorb the deleted or consolidated material.

Then assign each page an outcome: protect, consolidate, refresh, or retire.

For pages you remove, decide whether they require redirects.

For consolidation candidates, identify the surviving resource before dismantling the weaker pages.

Then watch what actually happens.

Did the surviving page gain impressions?

Did previously fragmented rankings consolidate?

Did AI systems begin citing the stronger resource?

Did conversions move?

Did a competitor replace you?

Or did you remove 100 URLs and discover that almost nothing changed?

That last result is not failure. If 100 pages disappear and neither buyers, search engines nor AI systems appear to miss them, you have learned something rather valuable about the library you were maintaining.

Publishing is cheap. Editorial judgment is becoming more valuable.

The seductive version of the QuickBooks story is still:

They deleted 2,000 articles, and traffic increased.

The more important lesson is that QuickBooks stopped assuming accumulated output was automatically an asset.

That’s more important now because AI can help your team create another article before anyone has finished asking whether the article should exist.

That changes what a sophisticated content operation looks like.

The advantage is no longer simply having the machinery to publish at scale.

It lies in being able to make stronger editorial decisions; knowing when one strong answer should replace four mediocre ones, recognizing when a ranking page is attracting an audience you no longer want, protecting obscure content that influences valuable buyer decisions, and being willing to retire work that has outlived the strategy that created it.

Expanding the library is now within everyone's reach. The real challenge lies in discerning what truly merits a place within it.

Jason Patel
Co-founder & CEO, Open Forge AI
AI Agents That Grow Your AI Brand Visibility

If you liked this newsletter, feel free to forward to your friends in marketing and AI.

If you got this from a friend, you can sign up to get the newsletter: https://aen.beehiiv.com/subscribe