Conventional GEO advice often treats owned content as the main lever. Two structured experiments tracked across 775 citation events suggest the opposite: the sources that cite you may matter more than the pages you control.
Both tests ran on 15 commercial-intent keywords. The first was a multi-month engagement for an existing brand across ChatGPT, Claude, Gemini and Perplexity. The second was a 30-day cold start for a SaaS link building agency with no measurable AI presence, adding Google AI Mode and Grok.
Earned placements dominated the citation mix
In the first experiment, listicles produced 72.4% of citations and PR produced 24.1%. Guest posts, the brand’s site and LinkedIn made up the rest. A comprehensive listicle on Indeed SEO became the largest citation source with 190 mentions after the model had already surfaced that domain about 10 times before outreach.
The brand appeared for roughly 10 to 12 of 15 keywords and peaked at 37.01% visibility on April 29. ChatGPT led with 148 citations, followed by Claude (96), Gemini (87) and Perplexity (64).
The second experiment was sharper. Third-party listicles generated 85.8% of 437 source mentions. The brand’s self-published listicle contributed 14.0%, and PR only 0.2%. Three sources—Indie Hackers, Bruce Jones SEO and the owned listicle—accounted for roughly 78% of total source mentions. Gemini led with 104 appearances, followed by Google AI Mode (95) and Claude (59).
This doesn’t mean owned content is worthless. In the cold-start test, the owned listicle was the slowest source to be cited at 18 days and worked better as a foundation than as a growth engine.
Entity associations changed visibility fast
One striking finding appeared in both tests. In the first, a listicle that placed the brand alongside recognized experts like Lily Ray and Aleyda Solis lost visibility within days after those names were removed. In the second, the owned listicle jumped from 4 to 49 mentions after the team added leading competitors to make the page more comparative—a 12.25-fold increase.
The takeaway: AI models appear to evaluate the entities around a mention, not just the mention itself. A category resource that includes competitors can outperform a self-promotional page with the same domain, author and keyword target. SERP visibility still influenced LLM visibility, especially in ChatGPT when web search was used to resolve a query.
A practical GEO playbook
Based on the experiments, brands should prioritize these steps:
- Prospect from observed citations. Run target keywords through relevant AI platforms and rank the sources already being cited. Use that list for outreach, not just domain rating.
- Budget for maintenance. About half of the sources stopped being cited within 30 days in the first test.
- Match placement type to query intent. Software and tool queries favored review sites; service queries returned more listicles.
- Build exact-phrase assets. The exact phrase Best LLM SEO Consultant ranked while a similar phrase without a dedicated listicle did not. The second test repeated the pattern with Best SaaS Link Building Agency in USA.
- Put key answers early. Pages performed better when the answer appeared in the first 100 words, with a key takeaway block near the top.
The tests also pointed to on-page signals that tracked with stronger citation performance: visible FAQ blocks, self-contained sections, question-based headings and fresh publication dates.
Citations are not clicks
The most-cited source in the second test, Indie Hackers, had flat referral traffic. TechBullion produced fewer citations but increased sessions from one to 64. One low-volume Perplexity referral led to a paying customer. Treat citation volume as a signal, not the final business outcome.
As AI search matures, the old ranking mindset isn’t enough. Track which sources cite you, who appears alongside you, and how recently those associations were reinforced.
Source: Search Engine Land
