From Zero to 298 AI Mentions in 30 Days — 85.8% Came From Third-Party Placements

Summary: A newly published experiment tracked a SaaS brand that had never appeared in an AI answer for 30 days. As reported, appearances went from zero to 298, and of the 437 source mentions behind them, 85.8% came from third-party listicles. The listicle on the brand's own domain took 18 days to earn its first citation and ended at 14%. It is a single-operator test with a small sample, but it points the same way as the Ahrefs experiment we covered in July, where 94% of an established brand's new mentions came from third parties. This article lays out the figures from both experiments, states the caveats, and then turns the finding into a procedure a brand can follow.

---

TL;DR

• Third-party placements produced most of the exposure. In the 30-day cold-start test, the 437 source mentions split 85.8% third-party listicles, 14.0% owned listicle, 0.2% PR (experiment-reported).
• The owned listicle was a foundation, not a growth lever. It was the slowest placement to be cited, at 18 days. Once competitor comparisons were added, its mentions rose from 4 to 49.
• Engine distribution differed between experiments. ChatGPT led the consultant test with 148 citations; in the cold-start test Gemini (104) and AI Mode (95) led while ChatGPT logged 32. One engine's result does not describe the others.

What the two experiments measured

Both experiments were published in Search Engine Land (Zeeshan Yaseen, 2026-09-14). The author is an AI search consultant who also runs a measurement tool, so there is a commercial interest. In both cases the author ran queries manually and logged citations, collecting 775 citation events in total.

Item	Experiment 1 — consultant brand	Experiment 2 — cold-start SaaS
Period	Several months (peak on April 29)	2026-05-30 to 06-28 (30 days)
Starting point	Some existing exposure	Zero appearances in the April 30 – May 29 baseline
Keywords	15 commercial-intent	15 commercial-intent
Platforms	ChatGPT · Claude · Gemini · Perplexity	Gemini · Google AI Mode · Claude · ChatGPT · Grok · Perplexity
Result	Peak keyword presence 37.01% (10–12 of 15 keywords)	298 appearances

The queries were unbranded commercial-intent questions — closer to "which service should I use for this?" than to "what is [brand]?". Why branded queries are the wrong baseline is covered in The branded-query baseline trap.

Experiment 1 — 72.4% of citations came from listicles

The first experiment tracked the author's own consultancy over several months. Broken down by source type, listicles dominated.

Source type	Share of citations
Listicles (third-party "best of" lists)	72.4%
PR (press and contributed articles)	24.1%
Guest posts · owned site · LinkedIn	Remainder

One article accounted for a disproportionate share. A single SEO listicle on Indeed produced 190 mentions, the largest single source. By platform, citations were ChatGPT 148, Claude 96, Gemini 87 and Perplexity 64.

From this the author concluded that publishing listicles on your own domain is a primary growth strategy. The second experiment led the author to revise that conclusion.

Experiment 2 — zero to 298 in 30 days, 85.8% third-party

The second experiment used stricter conditions. The author picked a SaaS brand with no AI presence, confirmed a zero baseline from April 30 to May 29, then spent 30 days from May 30 placing the brand in third-party listicles, publishing an owned listicle and running PR in parallel.

Result — 298 appearances, by platform

Platform	Appearances (30 days)
Gemini	104
Google AI Mode	95
Claude	59
ChatGPT	32
Grok	4
Perplexity	4

This is the reverse of experiment 1. ChatGPT, which led there, logged 32 here, while the two Google surfaces produced 199 — two thirds of the total. The author explicitly flags that "AI Mode wasn't tracked in the first experiment, and the two niches weren't directly comparable", so the comparison should be read with care.

Source mix — the owned listicle was 14%

The 437 source mentions behind the appearances broke down as follows.

Source type	Share (of 437)
Third-party listicles	85.8%
Owned-domain listicle	14.0%
PR	0.2%

"Experiment two suggested that an owned listicle is more of a foundation than a primary growth engine." — Zeeshan Yaseen, Search Engine Land

The top three sources (Indie Hackers, Bruce Jones SEO and the owned listicle) produced 342 of the 437 mentions, or 78%. At the individual level, the Indie Hackers post went from 44 to 146 mentions, Bruce Jones SEO from 26 to 69, and TechBullion from 15 to 37.

Two facts about the owned listicle

The listicle on the brand's own domain took 18 days to be cited for the first time, the slowest of all placements. But after a June 23 update that added competitor comparisons, its mentions rose from 4 to 49, and daily visibility peaked at 95 on June 27. The reading is not that owned content is pointless, but that it is slow on its own and works when it takes a comparative form.

Did it turn into traffic?

During the experiment, 18.5% of new users arrived via referral traffic and 3.25% via the GA4 AI assistant channel, 21.5% combined. The author is careful to note that citations and clicks are not the same thing: a single visitor from Perplexity — the platform with the fewest appearances, at 4 — became a paying customer, and there was no guarantee that the highest-volume platforms converted best.

The conditions to attach before using these figures

This is a small, uncontrolled experiment. The author's own caveats come first.

• A single operator, manual measurement, 15 keywords. The sample is orders of magnitude smaller than the 102,025-response study we reviewed in Brand tier decides how often AI names you.
• The two experiments used different niches, and experiment 1 did not track AI Mode, so the platform distributions cannot be compared directly.
• The author states the method "improved the hit rate, but it didn't guarantee citations".
• Citations decayed quickly. The author writes that checking only once, one week after publication, is not a reliable measurement approach. Our own measurement found run-to-run variance larger than the month-over-month change.
• The author sells consulting and a measurement tool. Read every figure as experiment-reported.

What makes the experiment useful despite this is that its direction agrees with an unrelated dataset. In the Ahrefs experiment we covered in July, 94% of an established brand's new mentions came from third-party content (self-promotion listicle analysis). Two experiments with entirely different samples and methods reached the same conclusion: most exposure comes from pages you do not own.

Engines differ — set against our Korean measurement

In the cold-start test, ChatGPT logged 32 appearances, less than a third of Gemini's. When RanketAI measured 12 Korean B2B SaaS brands in July 2026, the ChatGPT mention rate was 33%, less than half of Perplexity and Gemini at 83% each (Korean B2B SaaS benchmark).

The two datasets differ in sample, queries and period, so the numbers are not directly comparable. The common point is clear, though: ChatGPT tends to name new and small brands later than the other engines in both. Appearing in one engine does not imply appearing in another, and because each engine cites different sources, the third-party placements worth securing have to be checked per engine.

The procedure to follow

"Secure third-party placements" is not an actionable sentence on its own. What the experiment shows is that observing which sources are actually cited and getting into those beats adding more owned pages. The procedure has four steps.

1. Identify — collect the sources cited only for competitors on your target questions, recorded per engine.
2. Classify — sort them by how reachable they are. Below is the result of one RanketAI comparison run in the Korean AI marketing SaaS category in July 2026, classifying 15 such sources.
3. Act — do something different for each class.
4. Re-measure — not a single check one week later, but several runs read as a trend.

Class	Share (of 15)	Examples	What to do
Directly publishable (UGC)	3 (20%)	Naver Blog · Brunch · Wikidocs	Create an account and publish a rewritten version adapted to the channel. Zero cost, can start this week
Press or contributed-article route	5 (33%)	Business press · startup media · industry blogs	Package a launch or a research result as a press release or a contributed piece. As with the Indeed listicle in experiment 1, one article can produce hundreds of mentions
Structurally inaccessible	7 (47%)	Competitors' corporate blogs · public institutions	Drop from the target list

This is a single measurement in a single category, so the ratios cannot be generalized. The direction is what matters: roughly half of a gap list cannot be entered at all, and without classification you end up spending effort in the wrong places. The identification method and the measured case are in Source gap analysis. To see which sources cite only your competitors, per engine, use Competitor comparison; for per-engine mention and citation status, use AI brand visibility analysis.

What to do with the owned listicle

Keep it. In experiment 2 it was slow, but its mentions rose twelvefold once competitor comparisons were added. What worked on the owned domain was not a "we are number one" list but a table comparing competitors and the brand on the same criteria. That an owned list ranking your own product first on a broad category query risks being cited without being mentioned is something we confirmed in the self-promotion listicle analysis.

Frequently asked questions

Which comes first — a listicle on our own blog or getting into third-party listicles?

Third-party. In experiment 2, third-party listicles produced 85.8% of mentions and the owned listicle 14%. But this is a question of order, not either/or. Build the owned listicle in a competitor-comparison format as the foundation, and put the effort to grow exposure into third-party placements.

We have a small budget. Where do we start?

In order of cost, starting at zero: ① check which sources are cited only for competitors on your target questions, per engine; ② start with the directly publishable UGC channels (Naver Blog, Brunch and similar) among them; ③ bundle the press and contributed-article routes with a launch or a research result when you have one. In our measurement, UGC channels were 20% of the gap at zero cost.

298 in 30 days — can we expect a month too?

No. This was one person running 15 keywords, and the author states that citations were not guaranteed. Speed depends on category, engine and competitive intensity. The safe reading is that a brand with zero exposure can start appearing within a month through third-party placements; the actual pace has to be measured in your own category.

If citations go up, does traffic follow?

Not one-to-one. During the experiment 21.5% of new users arrived through referral and AI channels, but the paying customer came from Perplexity, the platform with the fewest appearances. Use citations as a leading indicator and judge outcomes against traffic and conversion data.

Is checking one week after publishing enough?

No. The author states that citations decayed quickly and that a single check one week after publication is not reliable. Collect several runs under the same conditions and read the trend; we have our own case where run-to-run variance exceeded the monthly change.

Takeaways

The contribution of this experiment is not the number 298 but the structure it shows: even for a brand starting from zero, most growth came from third-party placements. The author revised the first experiment's conclusion — that an owned listicle is a primary growth strategy — on the strength of the second. It is a small experiment by an author with a commercial interest, but it points the same way as the Ahrefs experiment with a different sample and method.

The actionable conclusion: build the owned listicle in a competitor-comparison format as the foundation, then identify and classify the third-party sources that cite only your competitors, per engine, and enter the reachable ones first. And read the results as a trend across several runs, not from a single check.

Related reading

• 44% of AI-cited sources are "best" listicles — the Ahrefs experiment, 94% of new mentions from third parties.
• Source gap analysis — finding the sources that cite only your competitors — the identify, classify, act, re-measure procedure with a measured case.
• Brand tier decides how often AI names you — how brand size structures appearance rates.
• Korean B2B SaaS AI visibility benchmark — the basis for the engine gap.
• The month-over-month illusion — run-to-run variance and measurement design.
• What is GEO (Generative Engine Optimization)? — definition and verified optimization methods.