10 min read
Content is worth creating only if it answers a real, specific question that nobody has answered well yet, at the moment a buyer actually has that question. If a topic doesn't clear that bar, publishing it won't produce traffic, trust, or pipeline; it will just add another page to a pile that search engines and AI systems increasingly route around. The practical shift this demands: stop asking "will this rank," and start asking "does this fill a gap that a person, or the AI summarizing for that person, currently has no good source for."
That sounds like a small reframe. It isn't. It changes what you publish, what you skip, and what you retire.
Why "publish more" stopped working even before AI answers showed up
The old logic of content marketing was simple: write about a topic, rank for it, collect clicks. That logic was already breaking down.
A 2024 study analyzing 416,386 clicks and 31.6 million impressions found that the #1 organic result on Google converts only about 9.28% of impressions into a click. Position 2 gets roughly 5.82%, and position 3 drops to about 3.11% — all lower than the click-through rates reported in older academic and industry benchmarks (Strzelecki & Miklošík, 2024). Even ranking well no longer guarantees a visit. That figure is one dataset, not a universal rule, and it varies by device and query type, but the direction is unambiguous: the payoff for simply ranking has been shrinking for years.
The reason isn't just more competition. It's that the search results page increasingly answers the question itself. A causal-inference study of 477,485 real search sessions found that when Google shows a direct answer or featured snippet on the results page, people click through to third-party websites less and spend more time on the results page instead (Gleason, Hu, Robertson & Wilson, 2023). This wasn't AI. It was 2019-era snippets and knowledge panels. But it's the exact mechanism that AI Overviews and chat-based answers have scaled up: when the platform satisfies the need on the spot, the click to your site simply doesn't happen, no matter how good your page would have been.
A more recent, still-early study of Google's AI Overviews (branded SGE at the time) argues this is accelerating zero-click search and pushing the discipline from "rank to be clicked" toward "get cited as a trusted source inside the generated answer" — a framing the researchers call Generative Engine Optimization, or GEO (Rajpoot & Raghuvanshi, 2026). This paper is very new and not yet independently replicated, so treat GEO as a plausible, well-reasoned direction rather than settled law. But it's the logical continuation of a trend that was already measurable years before generative AI existed. Nothing sudden broke. The same mechanism just got bigger.
None of this means content stopped mattering. It means the way content produces value changed: from being ranked and clicked, to being selected and cited as the raw material for someone else's answer.
Most of the decision already happens before anyone visits your site
There's a second, independent reason volume-for-volume's-sake content fails: most buyers have made significant progress on a decision before they ever talk to you.
Gartner data cited in a 2021 peer-reviewed B2B marketing paper found that buyers spend only about 17% of their total buying-process time meeting with potential suppliers across all channels combined (Paesbrugghe, Vuori & Kock, 2021). Separately, academic research on B2B buying behavior states that more than half of the B2B buying process happens through self-directed online research, and by the time a buyer contacts a seller, the decision is often already in an advanced stage (Marvasti et al., 2019). These are two different studies from two different years, not one precise statistic, but they point the same direction: the persuasion work has to happen in content buyers find and read on their own, long before a sales conversation exists to reinforce it.
Put those two forces together and the test for "worth creating" becomes obvious. Content earns its existence by resolving a real question at the exact point a self-directed buyer, or the AI system summarizing on their behalf, needs it answered — not by adding another variation of something already covered.
The three-way test: create, improve, or don't bother
Before writing anything new, audit what already exists for that topic — on your site, on competitors' sites, and inside the AI answers people are already getting. That audit resolves into three honest outcomes:
Create something new, if the question is real, asked often enough to matter, and genuinely unanswered — not "under-optimized," but actually missing a clear, direct, citable treatment anywhere.
Improve an existing page, if you already have content on the topic but it's vague, outdated, or too shallow to be the direct answer. Often the better move isn't a new article; it's rewriting the existing one so it actually resolves the question in the first few sentences, with specifics an AI system or a skimming reader can lift directly.
Skip it and pursue inclusion elsewhere, if the best possible answer already exists — on a review site, a comparison page, a forum thread, an industry publication — and your realistic goal is to be mentioned or linked from that source, not to out-rank it.
This is the same logic Tideflow AI's gap-analysis workflow uses: check what AI answer engines are already saying and citing for a brand's relevant questions, find what's missing or wrong, and only then decide whether the fix is a new article, a rewritten page, or a pitch for inclusion somewhere AI already trusts. The point of the check isn't to add more content. It's to publish the specific thing that's actually missing.
Applied honestly, this same test argues for pruning. If three pages on your site cover the same question with different titles, the AI systems and search engines evaluating your site have to guess which one is authoritative, and often guess wrong. Consolidating them into one clear, well-supported answer isn't busywork; it's the same gap-avoidance logic applied backward, to your own catalog instead of the market's.
What "worked" doesn't mean what it used to
Here's where teams get tripped up after publishing: they check traffic, see nothing moved, and conclude the content failed. That conclusion often isn't warranted, because traffic was never the only channel through which content can succeed anymore.
Three separate things can happen to a piece of content, and they need to be measured separately, not collapsed into one "did it work" verdict:
- Discovery — an AI crawler visited the page. This proves the content was found. It proves nothing else.
- Citation — the content (or a specific fact from it) got used inside an AI-generated answer. This is a meaningfully bigger deal than discovery, but it still doesn't guarantee anyone acted on it.
- Outcome — a person actually clicked through, arrived at your site, and did something you care about.
A bot visit in your server logs is evidence of discovery, not proof of citation, and citation is not proof of conversion. Treating any one of these as a stand-in for the others is a measurement mistake, and it's an easy one to make because discovery is the easiest signal to see. If you're evaluating whether a new piece of content earned its place, check all three, not just the one your analytics dashboard happens to surface first. Tideflow AI's own product documentation draws this line explicitly, separating crawler visits, citations, and conversions as distinct signals in its crawler observability tooling and conversion tracking, precisely because conflating them leads teams to overclaim success or wrongly declare failure.
A workable way to decide this week
Pick one topic you're considering writing about. Before drafting anything, do two things:
First, search for the exact question your buyer would ask (not the keyword — the actual question, in their words) using ChatGPT, Perplexity, or Google's AI-generated answer. Read what comes back. If the answer is already clear, well-sourced, and specific, your job isn't to write a competing version — it's to figure out whether you can get mentioned inside that existing answer, or whether the answer is quietly wrong or outdated in a way you can correct better than anyone else has.
Second, check your own site for anything that already half-covers the topic. If you find a page that's close but thin, rewrite it instead of starting fresh. A stronger single page beats two mediocre ones competing for the same authority.
Only write something entirely new when both checks come up empty: no one, including you, has a clear, specific, citable answer to the actual question your buyer has.
If you're doing this manually across dozens of topics, the audit itself becomes the bottleneck — which is exactly the workflow problem Tideflow AI's gap analysis is built to shorten, by comparing what AI answers already say against what your site covers and surfacing the create/improve/pursue-inclusion decision directly. But the underlying test doesn't require the tool. It requires asking, honestly, whether the piece you're about to write is filling a gap or just filling a calendar.
Frequently Asked Questions
Does declining click-through mean content marketing is no longer worth the investment?
No. The evidence shows clicks are harder to earn, not that content stopped mattering. Content can now succeed by being cited inside an AI-generated answer even when no one clicks through to the source site. The right response is reallocating effort toward content that resolves specific, unanswered questions clearly enough to be citable, not reducing investment across the board.
Will content that ranks well in traditional SEO automatically get cited by AI answer engines?
Not necessarily. Early research suggests AI citation selection weighs trustworthiness, source clarity, and direct-answer quality in ways that overlap with but aren't identical to classic ranking signals. This is an emerging area of study, not a proven equivalence, so a page that ranks well is a reasonable candidate for citation but not a guaranteed one.
How do I know if a crawler visiting my page means it's being used in AI answers?
It doesn't, on its own. A crawler visit only confirms the page was discovered. Whether it was actually cited in a generated answer, and whether that citation led anyone to your site, are separate questions that require separate checks — treating discovery as proof of impact is one of the most common measurement mistakes teams make right now.
Sources
- Strzelecki, A. & Miklošík, A., Device-dependent click-through rate estimation in Google organic search results based on clicks and impressions data (January 2024)
- Gleason, J.M., Hu, D., Robertson, R.E. & Wilson, C., Google the Gatekeeper: How Search Components Affect Clicks and Attention (June 2023)
- Rajpoot, P. & Raghuvanshi, R., An Exploratory Study on the Impact of Generative AI (Search Generative Experience - SGE) on Traditional SEO Strategies (April 2026)
- Paesbrugghe, B., Vuori, J. & Kock, H., Selling to enabled purchasers: the effect of perceived supply risk and profit impact on the buyer–seller interaction (August 2021)
- Marvasti, N.B., Huhtala, J-P., Yousefi, Z.R., Upreti, B.R., Malo, P. & Kaski, S., Is the potential B2B buyer interested or ready to buy?: Estimating the stage of the B2B buying behavior (August 2019)
- Tideflow AI, Tideflow AI — AI Visibility & Content Platform; Tideflow AI vs Profound: Which AI Visibility Platform Should You Choose?