Institution or Platform: What Health and Finance Citations Reveal About ChatGPT and Google AI Overviews

Last week's edition traced a single publisher's rise in Google AI Overviews back to a shift in what people were asking. This one goes a level deeper, into who each engine treats as an authority once the question is asked. And into how much room that leaves anyone else.

We pulled the cited-domain lists for healthcare and finance, the two categories Google's quality guidance holds to the highest standard. Both engines, in AI Hyper Cube. They produce different lists in the same category, in the same week.

What counts as a YMYL category

YMYL stands for Your Money or Your Life. It is Google's own term, set out in its Search Quality Rater Guidelines. It covers topics where low-quality content could plausibly harm a person's health, financial stability, safety or wellbeing. Pages in those categories are held to a higher standard than the rest of the web.

Source: Category definitions from Google's Search Quality Rater Guidelines. Coverage from BrightEdge AI Hyper Cube, week of 23 August 2026, United States.

We measured the first two. Legal guidance and news and civics are not in this dataset, so nothing here should be read as a claim about them. Both are on the list to run next.

The reason the distinction matters is that the guidance describes a standard rather than a source list. It does not name which domains an engine should cite in a sensitive category, and the two engines we measured have reached different answers. AI Hyper Cube reports the cited-domain list per engine, so you can check your own category rather than infer it from the guidance.

ChatGPT reaches for the institution

In tracked healthcare prompts, every one of ChatGPT's five most-cited domains is a government agency or a non-profit hospital system. Across the top ten, that holds for nine. YouTube does not appear at all.

Source: BrightEdge Insights from AI Hyper Cube. Citation presence, week of 23 August 2026, United States.

If your category is safety-adjacent, that is the citation diet you compete against on this engine.

What we analyzed

Two categories, two engines. For healthcare we took the five most-cited domains on each engine and their citation presence. Citation presence is the share of tracked prompts in a category where a domain appeared as a cited source. We also took the weekly healthcare series on Google, indexed so every source starts at the same point. For finance we took citation presence at each rank position within each engine's top five.

Each category is a separate prompt set. We compare figures within a category and engine, never across them. The prompt sets are different sizes, so comparing across them would compare the sets instead of the engines.

Data collected

Key finding

What the data shows. Measured: ChatGPT's healthcare top five is entirely government and hospital domains, with YouTube absent. In the same category on Google AI Overviews, YouTube stepped to roughly 1.8 times its June level in the week of 19 July, and held. Clinical partners stayed near baseline.

What we take from it. Inferred: the two engines are building different definitions of what counts as an authoritative source. The difference is large enough to change what a marketer should build. The data measures the lists. The read on what to build from them is ours.

Google reaches for the platform

The same category on Google looks nothing like ChatGPT's. In the week of 19 July, YouTube's citation presence in healthcare AI Overviews stepped up sharply and held there. NIH drifted below its June level.

Source: BrightEdge Insights from AI Hyper Cube. Google AI Overview citations, indexed so each series starts at 100 in the week of 7 June 2026, United States.

Brands with authoritative content still place. They now place alongside individual clinicians, museums and creators. One video surfacing in this set is a pathology museum explaining a condition with a specimen in hand. The asset was distinctive rather than institutional.

Finance splits on how much room is left

Finance shows the same disagreement in a different shape. On Google the leading source is cited in roughly half of all tracked prompts. The four below it are the largest sites in the category. On ChatGPT the same list is close to level.

Source: BrightEdge Insights from AI Hyper Cube. Citation presence as a share of tracked finance prompts, week of 23 August 2026, United States.

simplywall.st holds the fifth slot on ChatGPT. It is not a large publisher. It publishes community valuation narratives, where named contributors argue a fair value for a stock against analyst positions. No larger competitor in that list holds that view of a ticker.

Divide the leading source's presence by the fifth and you get a number worth running on your own category. Finance is 2.8x on Google and 1.7x on ChatGPT. A steep curve means an entrenched incumbent. A flat one means the category is shared.

How much the weighting changes by category

We added B2B technology as a control, because Google's strictest guidance does not cover it. The engine split appears there too, which suggests it tracks the engine rather than the sensitivity of the subject.

Source: BrightEdge Insights from AI Hyper Cube. Citation presence, week of 23 August 2026, United States.

Three categories is a pattern. A rule would need more. Run the same cut on your own category before you assume it behaves like any of these. Then compare it against your organic footprint in Data Cube X, to see whether the pages an engine cites are the ones already ranking.

What marketers need to know

1. Find the thing only you can say. Without a distinct point of view or data nobody else holds, selection is luck of the draw. simplywall.st takes a finance top-five slot on ChatGPT on community valuation narratives no larger competitor publishes. Establish what you or your audience know that the category cannot easily aggregate.

2. Read the citation list for your category, per engine. The winning source type is not a constant. Government and hospital domains take all five of ChatGPT's healthcare slots. Google AI Overviews puts YouTube above every clinical source in the same category. Pull the list before you decide what to build.

3. Check how much room the list leaves you. Divide the leading source's citation presence by the fifth. Finance runs 2.8x on Google and 1.7x on ChatGPT. Above three you are displacing an incumbent. Under two you are joining a rotation, which is a far cheaper fight.

4. Start with Google, and do not read format as a shortcut. Google AI Overviews carries the volume, so plan there first. Video and community formats reach places a long-form article does not, and that is a distribution decision. Foundational SEO remains the cost of entry to both surfaces.

Frequently asked questions

What is YMYL content?

YMYL stands for Your Money or Your Life. It is Google's term for topics where low-quality content could harm a person's health, financial stability, safety or wellbeing. Google's Search Quality Rater Guidelines group them into health and medical, financial advice, legal guidance, and news and civics.

What are the YMYL categories?

Four. Health and medical, covering diseases, treatments and mental health. Financial advice, covering investments, taxes, loans and banking. Legal guidance, covering rights and obligations. News and civics, covering current events, government policy and voting information.

Do Google AI Overviews and ChatGPT cite the same sources in YMYL categories?

Not in this dataset. In healthcare, ChatGPT's five most-cited domains are government agencies and hospital systems, and YouTube does not appear. Google AI Overviews places YouTube above every clinical source in the same category and the same week.

Can a smaller site get cited in a YMYL category?

In finance, yes. Citation presence on ChatGPT is spread almost evenly across the top five, and simplywall.st holds a position on community valuation narratives no larger competitor publishes. Concentration is far higher on Google AI Overviews, where the leading source appears in roughly half of tracked prompts.

How do you find out which sources get cited in your category?

Pull the cited-domain list for your category on each engine and compare them. BrightEdge AI Hyper Cube reports citation presence per engine. AI Catalyst tracks how your own share moves week over week. Data Cube X shows whether the cited pages are the ones already ranking in organic search.

Technical methodology

Key takeaways

Download the full report

The downloadable edition carries all three charts, the full cited-domain lists, the B2B technology control, and the methodology. It is the version to send to your search and content leads.

Find out what counts as authority in your category. Request a demo.

Explore more in our Weekly AI Search Insights series.

Pull your category's cited-domain list in AI Hyper Cube. Track your own share week over week in AI Catalyst. Compare it against your organic footprint in Data Cube X.