Much of today’s generative engine optimization advice rests on theory or a handful of screenshots. SEO consultant Zeeshan Yaseen ran two structured GEO experiments back to back, manually tracking 15 commercial-intent keywords in each and logging 775 citation events. One of his original conclusions did not survive the second test.
The first test, run over several months for an existing brand, covered ChatGPT, Claude, Gemini and Perplexity, with queries run with and without a VPN. The strategy placed listicles on sources LLMs already cited, plus the brand’s own site, supported by PR, guest posts and LinkedIn.
About half of the sources stopped being cited within 30 days. Authority followed a rough order — government and education, then news, industry sites and general sites — and PR amplified strong placements but could not rescue weak ones. One in-depth listicle outlasted five thin ones. A listicle that placed the brand next to Lily Ray and Aleyda Solis lost performance within days after those names were removed, suggesting models weigh surrounding entities.
On-page, answers in the first 100 words, a key-takeaway block near the top, visible FAQs, question-based headings and fresh dates all helped. Exact match still mattered: “Best LLM SEO Consultant” had a dedicated listicle and ranked, while “Best AI SEO Consultant” barely appeared.
The second test used a SaaS link building agency with zero AI presence and ran from May 30 to June 28 across six platforms, adding Google AI Mode and Grok. It produced 298 appearances: Gemini 104, AI Mode 95, Claude 59, ChatGPT 32, Grok 4 and Perplexity 4 — roughly two-thirds from Google’s surfaces.
When the owned listicle was rewritten to include leading competitors, its mentions rose from 4 to 49 — a 12.25-fold increase. It mirrored the first test’s peer-set effect in reverse. Yaseen revised his earlier advice: owned listicles are a foundation, not the main growth lever.
The most-cited source sent flat referral traffic, while a less-cited one grew sessions from 1 to 64. A single Perplexity referral — the lowest-volume platform — became a paying customer. Claude had fewer total appearances than Gemini but led or tied on five of the ten best-performing keywords, concentrating on high-intent queries.
Run target keywords through AI platforms before buying placements, and build outreach lists from sources already being cited. Budget for maintenance, not just publication. Manage who appears alongside you. Pair these findings with the difference between being cited and being recommended by AI and the 3,000-listicle study on counts and dates, then judge sources by visits and conversions rather than citation volume.
In the cold-start test, the owned listicle produced just 14.0% of mentions and took 18 days to be cited, while third-party listicles generated 85.8%.
In the first experiment, about half of the sources stopped being cited within 30 days, so GEO needs a maintenance budget.
Adding leading competitors to the owned listicle lifted its mentions from 4 to 49, suggesting AI favors pages that work as category resources.
Not necessarily. The most-cited source sent flat referral traffic, while the lowest-volume platform, Perplexity, produced a paying customer.
To apply what you just read to your own site, start with a free audit of where things are now.
A strategist replies within 24 hours on business days.
ChatGPT ads now run in Korea, but they don't change what ChatGPT says. Compare ChatGPT ads vs GEO on cost, placement and metrics, and see why reach limits make GEO essential.
With AI answers absorbing clicks, the top 100 media companies spent $113M on paid search in July, up 41% year over year. Forbes, CNN and USA Today still lost 20%+ of organic traffic. The loop this creates matters beyond publishing.
Google’s new Gemini 3.8 Live model now powers Search Live in the Google app, with web links and mid-conversation language switching. What it means for content owners.