How do I know whether content is creating demand or just capturing demand that already existed?
Short answer
Look at the language of the query bringing people to the page. If the query already names a solution category ("best CRM for agencies"), the buyer had decided to buy before they typed it, so the page is capturing demand. If the query only describes a problem ("cutting agent onboarding time"), the buyer hasn't connected their problem to a purchase yet, so the page has a chance to create demand. Confirm with branded search lift 4 to 12 weeks after publication, because creation content produces that lift and capture content usually doesn't.
Why the distinction changes how you budget
Demand capture pays back in months and stops working the moment a competitor builds something better on the same keyword. Demand creation pays back in quarters, has a much higher ceiling, and compounds because it changes what buyers believe about the problem before anyone starts comparing vendors.
Blending both into one "content revenue" number is why demand-creation content gets cut in Q1 and revenue softens in Q3, with nobody connecting the two events. Separate them in the report and the tradeoff becomes visible before you make the cut.
The query test that sorts every page in one pass
Export your top 100 pages from Search Console with their top query. Then sort each query by the language the buyer used to find you.
| Query language | Bucket | What it means |
|---|---|---|
| Names a solution category ("CRM software", "project management tool") | Capture | Buyer already decided to buy something |
| Names a competitor ("Asana alternative", "Notion vs Monday") | Competitor capture | Someone else created this demand |
| Describes a problem only ("reduce onboarding time", "team missing deadlines") | Creation candidate | Buyer hasn't connected problem to purchase |
| Names a concept you introduced ("pain point SEO", "product-led content") | Creation confirmed | Your framing entered their vocabulary |
Most B2B blogs come out roughly 55% capture, 25% competitor capture, and 20% creation candidates, and the surprise is how little of the thought leadership bucket lands in creation.
Competitor capture earns its own row because it's genuinely a third thing. You're harvesting demand a competitor created, which is profitable and worth funding, but it isn't your brand building a market. See How do I find non-brand demand we can realistically win?.
How to confirm a creation candidate actually created demand
The query test gives you candidates. Confirmation comes from what happened after publication.
Check branded search lift. Compare branded search volume for the 12 weeks after the page went live against the 12 weeks before. Creation content produces a measurable lift as readers internalise your framing and return searching your name. Capture content rarely does, because the buyer already knew you.
Check whether new queries appeared. If the page introduced a term, that language should surface as new queries over the following two quarters. No new query language means the framing didn't stick.
To isolate the lift from other causes like paid campaigns or a PR moment, see How do I test whether an organic lift was real and not seasonal or brand-driven?.
The AI-search adjustment most teams miss
Your query test will undercount demand creation, because a growing share of the buyers reading your framing never touch your site. According to our own state of AI search research, which analysed over 22,000 answers, Google AI Mode returns a citation-rich answer in 97.4% of matched B2B queries and Perplexity mentions the tracked brand in 41.7% of matched answers. A buyer reads your framing inside a ChatGPT answer, never clicks, then shows up two weeks later as branded search that Search Console credits to nothing. Pew Research found that when an AI summary appears, users click a source link in only about 1% of visits, which puts a number on how routinely that path replaces a click.
This matters more for creation than capture, because creation content is exactly what AI answers summarise well. So add a fourth signal: whether the page is cited on non-brand commercial prompts. A page cited where buyers describe a problem rather than name a category is creating demand at scale, whatever its click count says.
The compounding case is strong here. Our research found that once a brand is mentioned in an AI answer for a specific prompt, next-observation mention probability is 83.2% on ChatGPT, 83.3% on Perplexity, and 84.2% on Google AI Mode, versus roughly 10% after a miss. Creation content that earns a citation keeps earning it.
Run the classification
1. Export your top 100 pages with top query from Search Console. Last 12 months, non-branded queries only.
2. Sort every query into the four buckets. Capture, competitor capture, creation candidate, creation confirmed.
3. For each creation candidate, pull branded search 12 weeks pre and post publication. Flag any page with a lift above 5%.
4. Check which pages are cited by AI answers on problem-language prompts. These are creation wins your analytics can't see.
5. Report capture and creation as two separate revenue lines. Never one blended number.
Automate it in Analyze AI
Doing this once tells you where you stand. Running it monthly catches the drift that turns a creation-heavy library into a capture-heavy one without anyone deciding to.
GSC Keyword-Page Breakdown returns every page a query sends clicks to, which is how you detect the language shift that signals a page's classification changing. citation-magnets supplies the fourth signal by surfacing the pages AI surfaces cite most, which is the creation evidence Search Console structurally cannot give you.
The monthly agent:
Start (schedule, 1st of month) → GSC Top Pages for Site → Loop / For Each over the top 50 pages → GSC Page-Keyword Breakdown per page → GSC Keyword-Page Breakdown on the primary query → citation-magnets recipe → Citation Share → GA4 AI Traffic Overview → GSC Top Keywords for Site (branded regex, for the lift check) → workflow-memory (prior runs, to detect query-language change over time) → Code (classify each page into the four buckets by query language, correlate publication date against branded lift, and flag pages cited on problem-language prompts) → Excel export → Send Email to the content lead.
workflow-memory is what makes signal two workable at all. Detecting that new query language appeared requires knowing what the language was before, and an agent reading its own prior runs accumulates that history automatically rather than depending on someone having exported a query list a year ago.
FAQ
Related answers
- How much of our organic revenue is new demand versus branded demand capture?
- Is our non-brand visibility falling while branded traffic hides the loss?
Want to classify every page as creation or capture on live data? Start a free Analyze AI trial and run the monthly classification from one dashboard.