Owned, earned, and AI-cited: rethinking the marketing funnel for generative search in 2026

Key takeaways

• The classic owned/earned/paid split still exists, but a fourth layer, AI-cited, now decides whether any of it gets seen. Product pages are the single most-cited content type on ChatGPT Search, at roughly a third of all citations in July 2026, nearly double their share from March.
• Earned media isn't optional anymore. It's the trust signal AI engines lean on most for anything unbranded, and single-platform bets on it are fragile: Reddit's ChatGPT citation share cratered from about 3.8% to under 1% in a single week in August 2026.
• Different AI engines behave like different search engines. ChatGPT cites roughly 5 sources per response, AI Overviews and Perplexity cite about 10. Your content strategy needs to account for which "funnel" you're actually competing in.
• Paid is creeping into AI answers too: ChatGPT went from zero ads to averaging over 20% of citation-enabled responses carrying an ad within a few months.
• Tools like Promptwatch can show you which of your pages get cited, by which engine, and what to fix, closing the loop between visibility data and content action.

The funnel everyone learned is now half-wrong

For about fifteen years, marketers organized their world into paid, owned, and earned media. Paid was ads. Owned was your website, your blog, your product pages. Earned was press coverage, backlinks, reviews, word of mouth. You'd build owned assets, amplify with paid, and hope earned media validated the whole thing in Google's eyes.

That model isn't dead. But it's incomplete, because it never accounted for a fourth category that now sits on top of the other three and decides whether any of them get seen at all: AI-cited. Whether ChatGPT, Google AI Overviews, Perplexity, or Claude pulls your page, your press mention, or your Reddit thread into an answer is a separate battle from ranking on a SERP or landing a backlink. You can win owned and earned and still lose AI-cited, because the retrieval logic behind a generative answer doesn't work like a ranking algorithm optimizing for the best ten links. It works like a research assistant deciding which five to ten sources are worth quoting.

McKinsey estimates that half of consumers already use AI-powered search, and that up to $750 billion in US consumer spend will flow through AI-powered search by 2028. That's not a rounding error on the old funnel. That's a new front door, and most companies haven't rebuilt their house to face it.

Why "owned" content behaves differently in AI answers

Here's something that surprised me digging into the citation data: owned content, specifically product pages, is winning more real estate in ChatGPT's answers than it used to, not less. Promptwatch's tracking shows product pages went from about 18% of ChatGPT Search citations in March 2026 to roughly 32.8% by July, making them the single most-cited content type, well ahead of listicles (9.7%), news articles (5.2%), and how-tos (4.1%).

That's a big shift in what "owned content that works" means. It's not your homepage or your mission statement. It's the specific page with specs, live pricing, and availability, structured well enough that an LLM can lift a clean answer out of it without hallucinating. If your product pages read like marketing copy instead of documentation, you're leaving citations on the table.

At the same time, within July, listicles, how-tos, social posts, and comparison pages ate into that product-page dominance a bit, each growing a percentage point or two at product pages' expense. The lesson isn't "only build product pages." It's that the AI-cited layer rewards a mix: transparent product pages for owned, and comparison/how-to content for the earned-adjacent, discovery-stage queries.

Earned media just got more important, and more fragile

Muck Rack's research on generative engine optimization makes a claim worth sitting with: for branded prompts, more than half of the citations AI systems pull come from earned media, not the brand's own site. For unbranded, generic prompts, that share is even higher. If nobody outside your company is writing about you, an LLM answering "best project management tool for agencies" has almost nothing of yours to synthesize into the response.

That tracks with something PR agencies have been saying loudly this year. As one write-up on AI's effect on PR put it, AI doesn't learn from SEO or SEM, it learns from earned media, specifically press coverage, which has traditionally been PR's domain. That reframes PR from a reputation function into a visibility infrastructure function.

But here's the catch, and it's the part most "just get more press coverage" advice skips: earned channels can vanish overnight, and at least one already has, twice. Promptwatch's data shows Reddit held a steady 3.8% share of ChatGPT Search citations through late July and early August 2026, then collapsed to under 1% by August 14, an 86% relative drop, coinciding with a change in how ChatGPT Search fans out its queries. This wasn't the first time. In September 2025, Reddit had accounted for 10 to 14% of all ChatGPT citations before falling roughly 90% within days after Reddit changed its data-access terms.

That's the argument against betting your AI visibility on any single earned channel, no matter how dominant it looks this quarter. A platform's business decision, not your content quality, can zero out a third of your citation volume in a week. Promptwatch's report on Reddit's citation collapse frames the practical response plainly: content on domains you control is the only citation source no other company's decision can switch off.

Different AI engines run different funnels

A mistake I see constantly: treating "AI search" as one channel with one set of rules. It isn't. Promptwatch's data on average sources per response shows ChatGPT cites roughly 5 sources per web-search-triggered response, about half of a classic ten-blue-link Google page. Google AI Overviews and Perplexity both sit around 10 sources per response, a much more forgiving environment for a mid-authority or newer domain to get a slot. Microsoft Copilot, meanwhile, is wildly volatile, swinging from under 2 to nearly 17 sources per response within weeks, a sign Microsoft is still rebuilding its retrieval and attribution stack.

The practical read: ChatGPT is the fiercest competition per citation slot, because there are fewer slots to fight over. If your content strategy assumes "getting cited anywhere in AI search" is one goal, you're underweighting how much easier it is to earn a slot in AI Overviews or Perplexity than in ChatGPT.

The social platform breakdown tells a similar story. Reddit-collapse aside, ChatGPT is what Promptwatch calls a "Reddit specialist," citing it at roughly 5.19% even after some recovery, versus under 0.25% for any other social platform. AI Overviews and Grok lean on YouTube instead, at 4.08% and 4.85% respectively. Grok is the outlier that also gives real weight to Facebook and Instagram. There is no single social strategy for AI search, and building your earned-media plan around "we're big on Reddit" only helps you with one engine.

Paid is quietly re-entering the picture

The old owned/earned/paid triad assumed paid was straightforward: you buy placement. That's starting to happen inside AI answers too, and it's newer than most marketers realize. Promptwatch's ad-tracking data shows ChatGPT Search served zero ads until May 27, 2026. Within a few months, the ad rate climbed to averaging over 20% of citation-enabled responses over a 90-day window, with daily peaks near 44%. Most of those ad-bearing responses, about 73%, come from organic, non-branded prompts rather than brand-name searches.

ChatGPT's shopping features, product cards and price comparisons inside answers, are also being actively tuned, with usage rates that have visibly doubled overnight in past updates. If you sell anything, this is worth watching closely: it means there's now a paid layer sitting directly inside the answer, competing for the same attention as your organic citation.

What this means for how you plan content

Putting the pieces together, a 2026 content plan built for generative search needs to do four things the old funnel never asked for.

First, treat product and pricing pages as citation assets, not just conversion pages. Structured specs, transparent pricing, and live availability information are what gets lifted into an AI answer. Marketing fluff on a product page is now actively hurting your AI-cited visibility, not just your conversion rate.

Second, diversify your earned media across formats and platforms instead of concentrating on one. Original research, expert commentary, and independent journalist coverage carry weight that no single social platform can replace, and they can't disappear because a platform changed its terms of service.

Third, write for how AI engines actually search, not how humans type into Google. ChatGPT's query fanouts have gotten shorter and more keyword-like over 2026, dropping from roughly 117 characters in December 2025 to about 53 by April. Headings that read like short search queries, comparison pages, alternatives pages, and how-to clusters each represent a separate fanout sub-query and a separate citation opportunity.

Fourth, stop assuming one visibility number tells the whole story. Because ChatGPT, AI Overviews, Perplexity, and Copilot behave so differently, from source counts, to platform preferences, to how much you'll be competing with ads, you need visibility data broken out by engine, not a single blended score.

Promptwatch tracks exactly that kind of engine-by-engine breakdown, along with crawler logs showing when AI bots actually visit and read your pages, citation trends by content type, and Reddit/YouTube-specific reporting so you're not guessing which earned channel is actually driving your visibility this month.

Promptwatch

Track and optimize your brand visibility in AI search engines

View more

A practical checklist for the new funnel

• Audit your product and pricing pages for structured, factual clarity, not marketing language. This is now your top owned-content lever.
• Map which earned channels currently drive your citations, and build a contingency plan for the ones that could vanish (any single social platform, any single publication relationship).
• Segment your visibility tracking by AI engine. A citation win in AI Overviews and a citation win in ChatGPT Search represent different competitive dynamics and shouldn't be reported as one number.
• Build comparison, alternatives, and how-to content deliberately, since these formats are growing fastest in ChatGPT's citation mix and face less competition than product pages.
• Watch for ads and shopping features inside AI answers in your category. If ChatGPT is already running ads on your commercial prompts, your organic citation is now competing with a paid unit inside the same response.

If you're rebuilding a content or GEO strategy from scratch and want a broader look at what's out there beyond a single platform, the GEO software directory at bestgeosoftware.com is a reasonable place to compare options side by side. For agencies that want this handled end to end rather than run in-house, 1001 SEO Media combines technical SEO, content production, and generative engine optimization work under one roof, which is the kind of cross-functional setup this new funnel actually demands, since no single team, PR, SEO, or content, owns the whole picture anymore.

The bottom line

Owned, earned, and paid haven't disappeared. They've been absorbed into a bigger question: does an AI system trust your content enough to cite it, and does it survive long enough to matter once a platform changes its rules. The brands treating AI citation as a fourth, connective layer across their existing media mix, rather than a separate checkbox, are the ones building funnels that will still work in 2027.