Content Marketing After AI: What Is Still Worth Publishing
Table of Contents
Commodity information has stopped earning clicks, and it is not coming back. Definitions, generic how-to posts, and anything an assistant answers in one turn now get read inside the response rather than on your site. If your blog traffic has been sliding while impressions hold steady, that is the pattern you are looking at, and publishing more of the same will not reverse it.
What still earns attention is content marketing built on something a model cannot produce on its own: first-hand experience, data you own, a real opinion, local specificity, and judgement at the point of decision. There is also a second measure of value that most businesses are not tracking. A page can be quoted by AI systems hundreds of times a month while sending almost no visitors, and that invisible presence shapes buying decisions before anyone reaches your site. What follows covers which pages have stopped working, which ones survive, and how a content plan rebalances around both
What AI Took From the Blog
The losses in content marketing are concentrated in one type of page: content that answers a question completely, in a paragraph, with information available everywhere else. If a reader’s question is fully resolved by the first two sentences of an AI answer, the click has no job left to do.
Look at your own analytics and the pattern usually shows up as impressions holding steady while clicks fall away. The page is still being surfaced and read by machines. It just is not being visited. That gap between impressions and clicks is the clearest signal that a page has been absorbed rather than beaten by a competitor.
Anything built to capture that traffic through volume alone has lost its economics. A content audit framework applied honestly across an older blog will usually find that a third or more of the library falls into this category. Those pages are candidates for merging, retiring, or rebuilding with something the answer engines cannot supply themselves.
| Content type | What happened after AI answers | What to do with it now |
|---|---|---|
| Definitions and “what is” posts | Answered in the response, click rarely needed | Merge into a parent guide or retire |
| Generic step-by-step how-tos | Reproduced instantly with no visit | Rebuild around your process, costs, and failures |
| Listicles and round-ups | Summarised and stripped of context | Replace with decision criteria and trade-offs |
| Statistics round-ups without original data | Cited without attribution, low click value | Publish your own numbers instead |
| Service and decision-stage pages | Still visited, because buying needs proof | Deepen with pricing, process, and evidence |
| First-hand accounts and case studies | Increasingly quoted as source material | Publish more, with named authors |
The middle column matters more than the panic around it. Google’s own guidance on generative AI features states there are no special optimisations required for AI Overviews or AI Mode, because those features draw on the same ranking and quality systems as ordinary search. The rules did not change. The rewards moved.
Citations Are the Second Currency
A page can be quoted daily by AI systems while sending you almost no visitors, and that visibility still shapes buying decisions. It happens where you cannot see it, in a conversation between a prospect and an assistant, before anyone types your name into a search box.
Here is a real example from ProfileTree’s own reporting. In the Bing Webmaster Tools export for February 2026, one older guide on profiletree.com was credited with 17,613 citations in AI answers during the reporting period. In that same period, the page produced 39 organic clicks from Bing. Judged on traffic, it is a page you would delete. Judged on citation volume, it is one of the most-quoted assets on the domain.
The wider export tells the same story at scale: 1,199 URLs on the site were cited in AI answers during the period, 329 of them more than fifty times, and the majority recorded no matching organic traffic at all. Two currencies, running in parallel, measured in different places.
“The pages being quoted most often are rarely the pages driving the most traffic. When we started reading the citation reports alongside Search Console, we stopped deleting content that looked dead and started asking a better question: is this page feeding the answers our buyers are being given?” says Ciaran Connolly, founder of ProfileTree.
For a business owner, the practical consequence is that content value has to be assessed on two axes. A page that earns visits is doing recognisable work. A page that earns citations is doing invisible work: appearing as the source behind a recommendation, a comparison, or a cost estimate that a prospect reads before ever landing on your site. Understanding how AI search engines choose what to cite is now part of the content brief, not a separate technical exercise.
The uncomfortable part of content marketing after AI is that citation without attribution is normal. Your material can shape an answer with no link, no logo, and no visit. That is the cost of the format, and it does not make the work worthless. It makes brand recognition, named authorship, and distinctive claims more valuable, because those are the elements that survive summarisation.
Five Things a Model Cannot Generate for You
Every model can write a competent 1,500 words of content marketing on almost any topic. What none of them can do is originate the material that makes a piece worth reading. Five categories hold their value, and each maps to something your business already has.
First-hand experience
The specifics of doing the work: what a project actually cost, how long it took, what went wrong in month three, which supplier let you down. Published honestly, this is unrepeatable content. A guide to WordPress migration written by someone who has done forty of them reads differently from one assembled out of documentation, and readers can tell the difference within two paragraphs.
Proprietary data
Numbers only you hold. Client results, internal benchmarks, survey responses from your customer base, performance figures across your own channels. Original data earns citations because it cannot be sourced anywhere else. It also gives every other page on your site something to point at. If your business runs any kind of recurring reporting, you are sitting on publishable material.
Genuine opinion
A position someone could disagree with. Most published content avoids this, which is why most published content is interchangeable. Saying that a platform is wrong for a particular kind of business, or that a popular tactic wastes money for firms under a certain size, gives readers something to weigh. It also signals a human decision-maker behind the words, which matters more as AI content detection becomes routine in editorial review.
Local specificity
Regulation, funding schemes, market conditions, and buying behaviour vary by place. Content written for SMEs in Northern Ireland can reference Invest NI support, the realities of a small business community where reputation travels quickly, and search behaviour that differs from London or Dublin. Generic national content cannot compete with that on relevance, and the same logic drives AI-assisted local SEO for service businesses.
Decision-stage judgement
The advice that helps someone choose. Not what a service is, but whether it suits a business like theirs, at their size, with their budget, and what the alternatives cost. This is where buyers still click through, because they need a human answer to a question they cannot resolve alone. It is also the content that supports enquiries most directly, which is why content marketing services now start with the decision-stage pages rather than the top of the funnel.
How to Rebalance a Content Plan
Fewer pieces, deeper ones, tied to services, with named authors and evidence. That sentence covers most of what a content marketing plan needs to change, and the detail is in the sequencing.
Cut the publishing target first. A plan built around forty posts a year cannot produce work that carries first-hand evidence, because nobody has forty original things to say in twelve months. Six to ten substantial pieces, each 2,000 words or more, each carrying something specific, will outperform the larger number on every measure that pays.
Tie every piece to a service. Content marketing with no commercial destination generates traffic that goes nowhere. Each article should sit within reach of a page that sells something, connected by internal links placed early in the body rather than clustered at the bottom. This is the point where content and search engine optimisation stop being separate line items on a budget.
Put a name on it. Author credentials became a first-class ranking input in the February 2026 core update, with a dedicated Authors section added to Google’s Search Central documentation. Beyond ranking, a named author with a verifiable profile gives an AI system an entity to attach the claim to. Anonymous content contributes to answers without building anyone’s reputation.
Build evidence into the brief. Before writing starts, decide which number, project story, or original observation the piece will carry. If the answer is none, the piece does not need to exist yet. This single constraint removes most thin content before it is commissioned.
Refresh what earns citations. The pages already being quoted deserve maintenance ahead of new production. Update the figures, expand the sections that answer adjacent questions, and preserve the structures that make extraction easy: tables, numbered steps, and short direct answers under clear headings. Rewriting a well-cited page into flowing prose can quietly remove the thing that made it citable.
Give it eighteen months. Content built this way compounds slowly and holds value for years. Businesses expecting a quarterly return from a content programme are usually the ones who abandon it a month before it starts working, a pattern worth reading about in more depth in the future of content marketing.
Measuring Content When Clicks Stop Telling the Whole Story
Traffic reporting alone will now tell you a content marketing programme is failing when it is quietly succeeding. Four additional measures give a fuller picture.
Citation volume comes from Bing Webmaster Tools, which publishes AI page statistics per URL. It is the only widely available direct count of how often your pages feed AI answers, and it is free. Branded search volume in Search Console shows whether awareness is growing even where clicks are flat, and it was the single strongest resilience signal in network-wide analysis of the December 2025 update. Enquiry source data, gathered by asking every new lead how they came across you, catches the visits that analytics attributes to nothing. Query-level impressions show which questions your site is being surfaced for, whether or not anyone clicks.
Read together, these tell you something a traffic chart cannot: whether your content is present in the conversations where buying decisions form.
Where to Start If Your Chart Looks Like Everyone Else’s
The decline in blog traffic across the SME sector is not a failure of content marketing execution, and treating it as one leads to more of the content that stopped working. The businesses adapting well have made a smaller number of larger bets: fewer articles, more evidence, named people behind the writing, and a clear line from each piece to something they sell.
Start with a content marketing audit that sorts existing pages into three groups: quoted, visited, and neither. Keep and maintain the first two. Merge or retire the third. Then commission the next piece only when you can name the data or the experience it will carry. That constraint alone changes what gets published, which is the point.
Content Marketing After AI: Common Questions
These are the questions SME owners and marketing managers raise most often once the traffic charts stop behaving the way they used to.
Is blogging dead for small businesses?
No, but the type of blogging that worked between 2015 and 2023 has stopped paying. Publishing high volumes of short posts targeting informational keywords now produces impressions without clicks, because those questions get answered directly in AI responses and search features. What still works is a smaller number of substantial pieces built on first-hand experience, original data, and decision-stage advice. A trades business publishing four detailed articles a year about real jobs, costs, and problems encountered will generate more enquiries than one publishing forty generic posts. The change in content marketing is in the format and the standard, not the channel.
How do I know if AI systems are citing my content?
Bing Webmaster Tools provides an AI page statistics report showing citation counts per URL, which is the most direct measurement currently available to site owners at no cost. Google Search Console does not separate AI Overview appearances from ordinary impressions, so the signal there is indirect: watch for pages holding impressions while losing clicks. Beyond the tools, ask assistants the questions your customers ask and see which sources they name. Doing this monthly for ten target questions gives a usable picture of whether your business appears in the answers that matter, and which competitors are being named instead.
Should I stop writing content that only AI will read?
Content that feeds AI answers is doing commercial work, so the answer is no, provided the content connects to something you sell. A page quoted in answers about pricing, process, or supplier selection is influencing decisions before the prospect reaches your site. The pages worth stopping are the ones that neither get visited nor get cited, which is a different category. An audit that sorts pages by both measures will show the difference quickly, and it usually reveals that the two lists overlap less than expected.
How much content should a small business publish now?
Six to ten substantial pieces a year suits most SMEs, alongside quarterly refreshes of the pages already performing. The constraint is not budget but material: a business can only publish what it genuinely knows, and original experience accumulates at the pace of the work. Publishing more than that usually means writing about topics without first-hand knowledge, which produces exactly the content that has stopped earning attention. Businesses with active client work, regular data, or a strong point of view can sustain more. Most cannot, and pretending otherwise wastes the budget.
Does AI-assisted writing hurt rankings?
Google’s documented position is that content is assessed on quality and usefulness rather than production method, so using AI tools for research, structuring, or drafting is not itself a problem. What causes damage is publishing unreviewed output: generic phrasing, unverified statistics, no original material, and no editorial judgement applied. The practical standard is that a named person takes responsibility for accuracy and adds something the tool could not, whether that is data, experience, or opinion. Content that passes that test performs regardless of how the first draft was produced.
What should a content budget be spent on first?
Decision-stage pages, ahead of blog content. Service pages, pricing explanations, process descriptions, and case studies serve buyers who are close to enquiring, and they are the pages AI systems draw on when recommending suppliers. Once those are strong, move to a small number of deep guides carrying original data or first-hand accounts, each linked to the relevant service. Broad informational content marketing sits last in the queue now, and for many SMEs it does not justify a place in the plan at all.