EXPERIMENT

Can a New Domain Earn Brand Citations in ChatGPT?

I’ve seen enough SEO case studies to know how easy it is to choose the starting point after you already know the ending.

DankRank gives me a chance to do the opposite. The domain begins without inherited authority, an established backlink profile or years of branded searches behind it. If ChatGPT has no reason to know or cite the site at launch, that is exactly where the record should begin.

Experiment 001 follows the first 90 days and asks one question: can a new domain begin earning citations for non-branded prompts in ChatGPT?

I published the rules before Day 0. That way, the prompts, definitions and timing couldn’t be quietly rearranged after we saw the starting result.

  1. the prompt
  2. sources
  3. Citation

sources

PROOF / EXPERIMENT 001

EXPERIMENT RUNNING
  1. PROTOCOL PUBLISHED
  2. EXPERIMENT RUNNING

01

Experiment Status

Experiment running.

Day 0 was recorded on September 10, 2026 after the production domain was live, available to crawlers and documented in the launch visibility baseline.

Day 0 produced no observed DankRank citations, mentions or recommendations in the ChatGPT panel. That zero is the starting point, not a result to dress up as progress.

02

The Hypothesis

My working theory is that a new domain has a better chance of becoming useful to ChatGPT when its pages have clear jobs, the information is original, the brand is described consistently and the supporting evidence is easy to find.

That doesn’t give me a citation formula. I don’t know how long discovery will take, which pages might appear first or if DankRank will be cited during the 90-day window at all.

At this point, I genuinely don’t know which of those outcomes we’ll get.

03

What Counts as a Brand Citation

Seeing the name DankRank in an answer will be interesting, but it won’t automatically count as a citation.

Every observation will be recorded as one of the following:

  • Citation: A DankRank page appears as an identifiable source connected to the answer.
  • Mention: DankRank appears in the response without a cited DankRank page.
  • Recommendation: DankRank is suggested as an agency, provider or source when the prompt calls for that kind of recommendation.
  • No visibility: DankRank isn’t cited, mentioned or recommended.

Those definitions stay in place for the full experiment. A mention won’t become a citation later because the citation count looks better that way.

Every observation

  1. Citation
  2. Mention
  3. Recommendation
  4. No visibility

04

The Prompt Set

The core test contains 24 non-branded prompts divided into four documented groups. Some are commercial questions where an AI search agency could reasonably appear. Others relate to the Methods, Research and Resources we publish.

“What is DankRank?” isn’t a useful test of discovery because the prompt has already supplied the brand.

The same 24 prompts will be used at every scheduled checkpoint. New questions can be tested when there is a good reason, but they’ll be dated and kept outside the original score.

ChatGPT results can also change between sessions. We’ll record the testing environment and relevant context rather than presenting one answer as the answer every user receives.

05

The Site Will Change While the Test Is Running

DankRank can’t sit untouched for 90 days. Search engines need to crawl it, new evidence will be published and parts of the site may improve as we learn what isn’t working.

The experiment timeline will therefore include changes that could affect discovery, such as:

  • new pages going live;
  • substantial content revisions;
  • internal-link changes;
  • new PROOF or Research material;
  • external links or mentions we become aware of; and
  • technical changes affecting crawling, indexing or machine discovery.

If a citation appears soon after one of those events, the timing will be recorded. It may give us something worth investigating, but timing alone won’t prove what caused the result.

06

The First 30 Days Will Be Kept Relatively Quiet

I don’t want to launch the site, push a large promotion campaign immediately and then pretend we learned something clean about natural discovery.

Normal launch activity can still happen. The site will be submitted where appropriate, shared when relevant and allowed to earn legitimate attention. We won’t manufacture a large promotional spike solely to chase an early citation.

Any meaningful link, mention or promotional activity during those first 30 days will appear in the timeline.

07

What We Won’t Do

Buying citations would make the experiment worthless. So would publishing hundreds of thin pages whose only job is to repeat the DankRank name.

We also won’t remove prompts that continue returning nothing, count a plain mention as a citation or turn one unusual screenshot into a broad claim about visibility.

If the citation count is still zero at Day 90, zero is what we’ll publish.

08

Why Start With ChatGPT?

I chose ChatGPT for the first experiment because measuring every AI platform at once would muddy the result. Gemini, Perplexity, Copilot and Google’s AI experiences may discover and present sources differently, so combining them all into the first experiment would make a small result harder to interpret.

They’ll still be observed as part of DankRank’s wider AI visibility measurement. This particular test stays focused on ChatGPT.

For more about how citations work and why the same prompt can produce different answers, read the separate explanation.

READ HOW AI SEARCH CITATIONS WORK →

09

What Goes Into the Observation Log

For each run, we’ll record:

  • the prompt;
  • the date;
  • the ChatGPT environment and relevant session context;
  • whether DankRank was cited, mentioned, recommended or absent;
  • the DankRank URL shown, if there was one;
  • the other sources included in the answer;
  • any meaningful difference from the previous observation; and
  • known site changes or outside activity that could affect how we read the result.

A single screenshot is easy to overvalue. I want to see if the same thing happens again.

SEE HOW WE TRACK AI SEARCH VISIBILITY →

10

Results I’d Want to Investigate

The first genuine citation for a non-branded prompt would be the obvious milestone. What gets cited could be just as revealing.

ChatGPT might find an informational Method before it finds a commercial service page. DankRank could be mentioned repeatedly without any page being used as a source. A citation might appear once and vanish at the next checkpoint. The domain may remain invisible while other new sources enter the answers.

I don’t want to lump those outcomes together. A one-off citation, a recurring mention and 90 days of no visibility would tell us different things.

11

Day 7, 14, 30, 60 and 90 Updates

The scheduled checkpoints are Day 7, Day 14, Day 30, Day 60 and Day 90.

Each update will use the original prompt set and definitions. Earlier observations will remain visible, including zeroes, reversals and results that are too inconsistent to interpret confidently.

After Day 90, I’ll explain what happened and what I think the evidence supports. I’ll also be clear about anything the experiment still can’t answer.

12

The Commercial Question Is Separate

Even if DankRank earns citations during this experiment, that won’t prove we can make another brand appear in ChatGPT on command.

Client work involves a much wider search footprint. Citations matter, but so do mentions, recommendations, crawlability, organic visibility and the quality of the pages available for an AI system to use.

That work is explained on the AI Search service page.

13

Follow the Experiment From Zero

The DankRank Launch Visibility Baseline records the Day 0 starting point for Experiment 001. It is the reference we’ll use as the scheduled observations are added.

The rest of the build records, experiments and research are available in PROOF.