Belief 4 of 7 from the VizzEx AI Citation Study mid-study findings
In August, we made a call about one of our own clients, a cybersecurity firm. We’d checked about 100 of its buyers’ questions, and ChatGPT had cited it on 2. Its link profile was shrinking: 463 websites linking to it in January, 408 by August. No PR. No analysts. No trade press. Our read: this site would need outside validation before ChatGPT would cite it.
Then we checked every question its blog answers, about 300 more. ChatGPT was already citing it on a dozen of them. It has cited the firm on 10 to 15 of its questions at every check since.
Meanwhile, the most-linked site in our study, a famous publisher with 96,060 linking domains, hasn’t been cited by ChatGPT once.
We were wrong about authority. So, it turns out, is a lot of the advice about it.
The Belief: Authority and Amplification Are the Price of Admission
“You need a Reddit and community presence”, and its big brother: “Authority and amplification are the price of admission.”
Two widely shared frameworks say this.
Two Frameworks That Put Authority at the Center
- Möller’s Venn diagram puts AI visibility at the overlap of three circles: Authority × Relevance × Extractability. The post drew 449 reactions, and one line from its comment thread sums up how readers took it: “it’s nearly always authority that’s missing.”
- Informa TechTarget’s S.I.G.N.A.L. Index scores six dimensions. It treats Amplification (analysts, media, third-party endorsement) as the way AI comes to cite you. It ranks brands on a ladder of four archetypes:
| Archetype | Visibility the paper assigns it |
|---|---|
| Newcomer | 10% |
| Contributor | 50% |
| Referenced | 75% |
| Authority | 100% |
The Newcomer’s gap, in the paper’s words, is “lacking analyst relationships, original research, trade press, third-party validation.” Every archetype in the paper is labeled a simulated scenario.
So we tested the belief against live sites.
How we tested it
The VizzEx AI Citation Study tracks 10 live websites, including two famous marketing publishers we left untouched as a control group, on about 1,200 real buyer questions across Google AI Overviews, ChatGPT, Claude and Perplexity. Every answer is stored. One of the ten sites is ours, vizzex.ai.
For this belief we added one more measurement: each site’s link profile, from a link index. That’s how many other websites link to it, and how that changed over the year.
What High Authority Actually Bought in AI Citations
| Site | Websites linking to it | What the AI engines did |
|---|---|---|
| Famous marketing publisher A | 96,060 (about 1,500 lost during the study) | ChatGPT: 0 citations, every check |
| Famous marketing publisher B | 4,367 | ChatGPT: 0 citations, every check |
| Client site: cybersecurity | 408, down from 463 in January; no PR | ChatGPT cites it on 10 to 15 of its own questions at every check |
The two famous publishers aren’t invisible. Claude and Perplexity cite them often: 346 times in the first eight weeks of the study. But the answers named them in only 6 of those 346. They’re used as sources and almost never credited, and ChatGPT doesn’t use them at all. That gap between being sourced and being named is its own phenomenon — one we examine in depth in the finding that cited by Claude and Perplexity, almost never named comes down to the question type, not the brand.
The cybersecurity client is the opposite. Here’s ChatGPT, check by check, from the day its full set of questions entered the study: That pattern held even though the client’s traditional SEO signals were declining — a dynamic our control group study confirmed: Claude and Perplexity cite them often, yet strong SEO didn’t carry over into AI visibility for the high-authority sites.
| Check (day of study) | Questions ChatGPT cited the client on |
|---|---|
| Day 21 | 15 |
| Day 28 | 13 |
| Day 35 | 10 |
| Day 42 | 14 |
| Day 49 | 12 |
| Day 56 | 12 |
ChatGPT Citation Frequency: The Cybersecurity Client Week by Week
That includes the question about its own proprietary framework. ChatGPT has cited the client’s page for it at every one of those checks, on nothing but the client’s own word.
That’s a level, not a change. Most of those questions entered the study at day 21 and were cited the first time we checked them. We don’t know when ChatGPT started citing them. We know that it does, and keeps doing so.
The TechTarget Authority Ladder, Measured Against Live Data
TechTarget’s ladder puts the newcomer at the bottom with 10% visibility and the established authority at the top with 100%. We have a real site on each end.
| Newcomer (bottom rung) | Authority (top rung) | |
|---|---|---|
| The site | vizzex.ai, under nine months old | Famous publisher A |
| The paper’s description | lacks analyst relationships, original research, trade press, third-party validation | “always in the answer,” cited for foundational concepts |
| Websites linking to it | 86; 75 of them are press-release syndication; none earned independently | 96,060 |
| What we measured | 63 of 218 questions cited; 29 on ChatGPT; the highest naming rate in the study | ChatGPT: 0; cited by Claude and Perplexity, almost never named |
One more fact about the top rung: one of the famous publisher’s signature ideas now turns up in AI answers almost word for word, with no link and no credit. The answer has absorbed it.
On this evidence, the ladder looks upside down.
A disclosure about our own site. VizzEx issued three press releases this year. The August one, when VizzEx was listed in the WordPress plugin directory, came out in the same three weeks our citations jumped. So on our own site we can’t separate what the content did from what the press release did. That’s why the case for this belief rests on the client sites, not ours. Our site is here to show the Newcomer archetype is wrong. It doesn’t carry the claim.
What Happens Next: Tracking Original Research on Our Own Site
This study changes where vizzex.ai sits on the ladder. One of the Newcomer’s gaps in TechTarget’s paper is “original research,” and vizzex.ai now has some: this study. We’re publishing its findings here and in my LinkedIn newsletter, with no PR campaign behind them. So through the end of the study, we’re tracking what that does: whether our site is cited and named more on the questions this research answers, and whether the AI engines cite the newsletter articles themselves. On our own site it can’t prove what authority does on its own, but it will show what publishing original research did. We’ll report it either way.
Möller’s Three Circles (Authority, Relevance, Extractability) Measured Against Live Data
| Möller’s circle | What our data shows |
|---|---|
| Authority | The site with the most authority in the study got nothing from ChatGPT. A client with a fraction of it, and falling, is cited every week. Authority doesn’t decide the citation. |
| Relevance | The specific questions a site’s buyers ask are where the treated sites win. That’s Belief 6 in this series. |
| Extractability | This is the leg you can control. Authority takes years to build and, in our data, didn’t buy the citation. How easily an engine can read and lift your content is something you can change this quarter. |
| What none of them covers | Getting named. All three circles are about getting cited. A brand can score perfectly on all three and stay an anonymous footnote forever. That’s Belief 2. |
What Our Data Actually Shows About Reddit and Community Presence
We’re not saying don’t build authority or community presence. It may augment your visibility. It just isn’t the requirement for getting cited. Our own site was being cited by AI engines when it was two months old, before our first press release.
What our data does show about Reddit is where it wins: the ground nobody can own. Head terms, broad topics, “best X” questions. And when Reddit is cited, Reddit holds the position, not the brand being discussed in the thread.
Several of the studies recommending a Reddit playbook measure visibility with keyword-index tools. That kind of index shows our cybersecurity client’s ChatGPT citations as zero. Know what a study measured before you budget from it.
Verdict: Did not survive
Outside authority is not the price of admission.
The site with the most other sites pointing at it got nothing from ChatGPT. One with a fraction of that, and falling, is cited on a dozen of its own questions every time we check.
Rent a little presence. Own your positions.
Watch Belief 4: Mid-Study Findings Webinar Segment
Kim walked through this belief live in the Mid-Study Findings webinar, including the ladder, measured. Watch the Belief 4 segment here, or the full hour for all seven beliefs.
What You Can Do Today to Improve AI Citation Without Building Authority
1. If there’s an “AI authority” line in your budget, ask what it’s measured against. Analysts, placements and PR can still be worth doing. Just ask what result they’ll be measured on, and whether that measurement can see AI citations at all.
2. Look at what your own site says that nobody else can. Your proprietary framework, a customer story only you can tell, the specific questions your buyers ask before they hire you. In our data, that’s the ground ChatGPT cites a small site on, every week.
3. Check which of your tracked questions are rented ground. If Reddit, YouTube or Wikipedia hold the answer to a question, nobody owns it. Put your effort into the specific questions where the answer can only come from you.
What’s next
Belief 4 is one of seven things everyone says about AI search that we put on a scoreboard. Still to come:
- whether citation volatility is random — and how AI citation wins decay over time;
- whether head terms are the prize; If you want to get ahead of that one: head terms aren’t the prize in AI search — your buyer’s specific questions are, and domain authority doesn’t open that ground for you either.
- whether more content is the answer.
The study ends October 17. At the end we’ll re-score every row on the board against the final data, including the predictions we filed at mid-study, whichever way they go.
Kim Albee is co-founder of VizzEx with Carolyn Holzman. The VizzEx AI Citation Study is a controlled, 84-day study of how Google AI Overviews, ChatGPT, Claude and Perplexity choose, keep and name their sources.
Frequently Asked Questions
Does domain authority or the number of linking domains affect whether ChatGPT cites your site?
The site with the most other sites pointing at it got nothing from ChatGPT. One with a fraction of that, and falling, is cited on a dozen of its own questions every time we check. The two famous publishers aren't invisible. Claude and Perplexity cite them often: 346 times in the first eight weeks of the study. But ChatGPT doesn't use them at all.
Do you need analyst relationships, trade press, or third-party validation to get cited by AI search engines?
Outside authority is not the price of admission. Our own site was being cited by AI engines when it was two months old, before our first press release. It just isn't the requirement for getting cited.
What does Möller's Authority × Relevance × Extractability framework get wrong about AI citations?
The site with the most authority in the study got nothing from ChatGPT. A client with a fraction of it, and falling, is cited every week. Authority doesn't decide the citation. What none of them covers is getting named — all three circles are about getting cited. A brand can score perfectly on all three and stay an anonymous footnote forever.
Where does Reddit actually win in AI search visibility, and does a Reddit strategy help brands get cited?
What our data does show about Reddit is where it wins: the ground nobody can own. Head terms, broad topics, 'best X' questions. And when Reddit is cited, Reddit holds the position, not the brand being discussed in the thread.
Which factor in AI citation frameworks is most actionable for marketers right now?
Extractability is the leg you can control. Authority takes years to build and, in our data, didn't buy the citation. How easily an engine can read and lift your content is something you can change this quarter.