Two GEO Experiments: Earned Placements Beat Owned Listicles for AI Visibility

Most GEO advice is built on theory, isolated screenshots, or a single campaign. On 14 September 2026, Search Engine Land published something rarer: two structured experiments run back to back, with manually tracked results across six AI platforms.

The conclusion challenges a common playbook: an owned listicle is a foundation, not a growth engine. Earned placements on third-party sites generated most of the LLM visibility. The owned listicle was the slowest source to earn citations and contributed only 14% of mentions.

LLM visibility depends more on which sources mention you than on which pages you control.

What was tested

Zeeshan Yaseen's Search Engine Land article documents two sequential experiments tracking AI visibility for a SaaS agency brand across commercial-intent buyer queries.

The experiments tracked whether the brand appeared in AI answers, which sources were cited, and how visibility changed after specific GEO interventions — publishing an owned listicle, earning third-party placements, and reinforcing off-site associations.

This is not a 10,000-page corpus study. It is a controlled, longitudinal test on a defined keyword set with manual verification. That limits generalizability but increases actionability: you can replicate the design on your own prompt panel.

Finding 1: SERP visibility still influences LLM visibility — especially in ChatGPT

When ChatGPT relied on web search to resolve a query, brands absent from retrieved results were also absent from the answer. Search Engine Land's experiments confirmed the direction that Eugen Ullrich's retrieval-to-citation funnel analysis measured independently: ChatGPT retrieves broadly but cites sparingly (~41% citation rate), while Google AI Overview cites ~77% of what it retrieves.

The practical rule: you cannot be cited if you are not retrieved. For ChatGPT specifically, ranking in the search results the model pulls from remains a gate.

But retrieval is necessary, not sufficient. Experiment 2 showed that showing up in search results did not guarantee citation — and that off-site associations often drove mentions even when owned pages were retrieved.

Finding 2: Earned placements generated most visibility

The headline finding from experiment 2: earned placements on third-party sites generated the majority of LLM mentions. The owned listicle — initially recommended after experiment 1 because many top-ranking brands appeared to benefit from their own comparison content — turned out to be the slowest source to earn citations and contributed only 14% of total mentions.

This aligns with the broader data:

• Muck Rack's Generative Pulse found 84% of AI citations come from earned media.
• Indexably's study found domain factors explain 77% of predictive importance for AI citations — page optimization only lifts the top authority quartile.
• AEO Labs' cross-vertical data shows brand-owned sites account for only 2–6% of what AI cites.

An owned listicle can rank. It can even be retrieved. But AI engines disproportionately cite and mention brands through third-party associations — listicles on publisher sites, review platforms, community threads, and editorial roundups.

Finding 3: LLM visibility is an associations game, not a pages game

Both experiments pointed to the same mechanism: LLM visibility depends on which sources mention you, who appears alongside you, and how recently those relationships were reinforced.

Traditional SEO optimizes pages. GEO must optimize relationships — the network of third-party pages that describe, compare, recommend, or review your brand.

The experiment's main disagreement was about the role of owned content. Experiment 1 suggested publishing an owned listicle because top-ranking competitors had them. Experiment 2 showed the owned listicle was "more of a foundation than a primary growth engine" — necessary infrastructure, not the fastest path to AI mentions.

That matches how AI systems actually compose answers. They retrieve multiple sources, synthesize a recommendation, and cite the sources they trust. A brand's own "Best X Tools" page is one source among many — and often not the one the model selects, especially for commercial queries where Google cites YouTube, Reddit, and aggregators instead of vendor sites.

Finding 4: Platform coverage matters — one engine is not enough

Experiment 2 expanded to six platforms including Google AI Mode and Grok. Visibility was not uniform:

• A placement that moved ChatGPT did not automatically move Gemini or Perplexity.
• Foglift's Q3 benchmark found 75% of top-25 cited domains were exclusive to one engine.
• Rankability's August 2026 analysis found no pair of AI platforms shared more than 24.1% of cited pages.

The experiments reinforce a rule Obsurfable has tracked across the corpus: monitor the platforms your buyers actually use, not a single "AI visibility score."

What this changes about GEO roadmaps

Deprioritize: owned listicle as primary growth lever

Publishing a "Best [Category] Tools" post on your own blog is still worth doing — for SEO, for sales enablement, for entity clarity. But do not expect it to be the fastest or largest driver of AI mentions. Budget it as 14% infrastructure, not 80% strategy.

Prioritize: earned placements on surfaces AI already cites

1. Editorial listicles and roundups on publishers AI cites in your category (Forbes, Zapier-style comparisons, vertical trade press).
2. Review platforms — G2, Capterra, TrustRadius for B2B; category-specific directories for other verticals.
3. Community presence — Reddit, niche forums, YouTube reviews. Not astroturfing; genuine participation where buyers ask questions.
4. Comparison and integration pages on partner sites — co-marketing that creates third-party pages mentioning your brand.

Measure: mentions, citations, and sources separately

The experiments tracked whether the brand was mentioned in the answer and which sources were cited. Those diverge — a pattern AuspiaAI documented in September 2026 across 20 prompts on three surfaces, where ChatGPT named vendors on 13 of 20 prompts but cited sources on only 8.

Track all three:

Sequence: retrieval first, then distribution, then on-page

1. Can AI find you? Technical crawlability, index presence, basic ranking for category queries.
2. Do trusted third parties mention you? Earned placements, reviews, community presence.
3. Is your owned content ready to compound? Answer-first structure, extractable formatting — especially if you are in the top authority quartile.

Reversing that sequence — leading with an owned listicle rewrite before distribution — matches what experiment 2 found least effective.

Practical playbook

Week 1–2: Baseline

1. Define 15–25 commercial-intent buyer prompts (the experiment used 15).
2. Run them across ChatGPT, Perplexity, Gemini, and Google AI Overviews/AI Mode.
3. Log: brand mentioned? domain cited? which third-party sources cited?
4. Identify the source gap — publishers and platforms that cite competitors but not you.

Week 3–6: Earned placement sprint

1. Pitch 3–5 editorial listicles or comparison features on domains AI already cites for your prompts.
2. Optimize existing G2/Capterra/review profiles — AI reads these as consensus signals.
3. Seed genuine community answers on Reddit or niche forums for BOFU queries (see Reddit citation concentration).

Week 7–8: Owned foundation (not sprint)

1. Publish or refresh one owned comparison/listicle page — for SEO and entity clarity, not as the primary AI growth bet.
2. Structure it answer-first with extractable headings and comparison tables.
3. Re-run the prompt panel. Expect slower movement than earned placements.

Ongoing: Monthly re-measurement

Repeat the prompt panel monthly. AI visibility is volatile — only 30% of brands stay visible back-to-back. One experiment cycle is a starting point, not a verdict.

How Obsurfable fits

Obsurfable is built for the measurement design these experiments used manually: defined buyer prompts, repeated runs across AI engines, tracking mentions, citations, and competitor sources over time.

When an earned placement lands, Obsurfable shows whether it moved mention rate on the prompts you care about — and whether a competitor's third-party coverage still dominates. When an owned listicle produces only 14% of visibility, the data makes that failure visible early.

The Visibility Director identifies which third-party source gaps to close and drafts outreach or content aimed at the surfaces AI actually cites in your category.

FAQ

Should I stop publishing owned listicles entirely?

No. They are foundation — SEO value, sales collateral, entity signals. Just do not treat them as the primary AI visibility growth lever. Lead with earned placements; maintain owned content as infrastructure.

How long did the experiments run?

Experiment 1 ran May 30 – June 28, 2026 (~4 weeks). Experiment 2 was a follow-on with expanded platform coverage. AI visibility can shift faster or slower depending on category and authority — plan for monthly measurement, not weekly verdicts.

Does this apply to B2C as well as B2B SaaS?

The experiments used B2B SaaS buyer queries. The earned-media pattern holds across verticals — 84% earned media is cross-platform. B2C may see even heavier reliance on YouTube, Reddit, and review aggregators.

How does this relate to the CITECHOICE study published the same week?

CITECHOICE found structured formatting redistributes citation credit among retrieved sources but does not reliably increase admission. These experiments find earned placements drive admission (getting into the source set) more effectively than owned listicles. The two studies are complementary: distribute to get retrieved, then format to win allocation.

Bottom line

Two back-to-back GEO experiments published 14 September 2026 found that earned third-party placements generated most LLM visibility, while an owned listicle was the slowest source to cite at 14% of mentions. SERP visibility still gates ChatGPT retrieval, but associations — who mentions you, where, and how recently — drive the outcome.

Stop treating owned listicles as the primary GEO growth engine. Start treating third-party distribution as the citation layer — and measure it on a fixed buyer prompt panel, monthly, across every engine your audience uses.