How ChatGPT Does Question Fan-Out (15 Million Queries Studied)

We needed to see for ourselves how ChatGPT retrieves info—and what meaning for AI visibility—so we analyzed 66.6 million free, logged-out GPT-5.6 Luna responses and their 15 million fan-out queries from August and September.

Solely 14.4% of the 66.6 million ChatGPT responses we tracked triggered an online search.

When ChatGPT did search, it ran a median of about 1.6 fan-out queries for every response that searched.

Each internet search prices OpenAI cash, so it is smart for ChatGPT to skip one when it may possibly reply properly sufficient from what it already is aware of.

This suits with WordLift’s analysis which discovered that free-tier fashion fashions could lean extra on coaching knowledge to maintain prices down.

Which means that, for a lot of prompts, your probabilities of showing rely on what ChatGPT already is aware of about your model, somewhat than what it may possibly discover.

Of the 15 million fan-outs we analyzed, 6.8% used the website: operator to look a single web site.

And among the many ~160 most-searched websites, three in 4 of these searches focused a model website.

Retailers like Walmart, Costco and Amazon made up 11.9%, and Reddit and Fb simply 1.9% between them.

The phrases contained in the fan-outs present the identical give attention to official, first-hand info, with “official” and “value” the commonest after “2026”, and rising by just a few proportion factors month-over-month.

None of this can be a huge shock. If you need a product’s value or specs, a model website is the plain place to examine.

When a query has no single proper reply, that’s when ChatGPT is extra more likely to look past official sources.

As an example, in a separate take a look at of 20,000 US prompts, 84.4% of factual questions on pricing, product options or firm particulars cited no less than one official supply.

For comparability questions like “A vs B” or “which is cheaper”, that dropped to 68.3%.

The combo of sources modified too.

For factual questions, 63.4% of citations had been to official web sites.

For comparability questions, it was 46.4%, so most citations went to different websites.

So ChatGPT depends most on official sources for details, and leans extra on different websites when it has to weigh one possibility towards one other.

We’ll dive deeper into fan-out question language somewhat later.

The domains ChatGPT targets most with website: are huge client manufacturers.

These rankings are, after all, based mostly on our personal dataset and prompts, however with round 1 million website: queries behind them, they offer a powerful sense of which domains ChatGPT leans on most.

These are principally manufacturers that promote merchandise, so a model website is the pure supply for particulars like costs and specs.

Authorities domains additionally characteristic closely.

Gov.br, gov.uk, France’s service-public.fr, the NHTSA, the FDA and the SEC all sit within the prime 150, which matches the observations of Lily Ray, Founding father of Algorythmic, that ChatGPT favors official sources for YMYL questions.

Word additionally that Reddit continues to be the ninth most website: searched area.

Of the 17 publishers and platforms that at present have licensing offers with OpenAI, Reddit is the one one within the prime 150 most website: searched domains.

One doable motive is that ChatGPT already has its companion content material pre-loaded and able to go.

These website: searches present what ChatGPT appears up throughout question fan-out, however not what it finally ends up citing.

The cited domains inform a unique story. Associate publications are in all places.

In Model Radar, companion titles amongst ChatGPT’s 200 most-cited domains greater than doubled their mixed share of citations, from 7.5% in March to fifteen.8% in July—and nonetheless held 12.5% in late September.

The phrases ChatGPT provides to its fan-out queries trace on the web page varieties it desires to retrieve.

Between August and September, extra of ChatGPT’s fan-out queries had been about costs, product specs, and critiques, and barely fewer had been about definitions.

A acquire of 1 or 2 proportion factors can sound small, however these phrases all began from a low base, and their relative rise reveals starker development.

As an example, “Worth” went from 6.04% to eight.88% of fan-outs, a relative rise of practically half (47%) in a single month.

Including to this, “Evaluations” grew by a 3rd, and “specs” entered the highest 50 for the primary time.

On the flip facet, “That means” and “definition” every fell by a couple of fifth, although from small shares, in order that’s seemingly a dip somewhat than a pattern.

Alongside the fan-out knowledge, we checked out which websites ChatGPT cites.

This attracts on a separate dataset: Model Radar’s ChatGPT citations within the US from March to September 2026, overlaying 2.4 to three.8 million solutions a month.

Snippet Digital discovered that Reddit citations in ChatGPT dropped 86% in August, and knowledge shared by Aleyda Solís, founding father of Orainti, confirmed social, reference, and information websites shedding essentially the most quotation share between July and August.

Our Model Radar knowledge reveals the identical drop throughout Reddit, Wikipedia, and YouTube.

In March, about 1 in each 5 ChatGPT solutions cited Reddit, however by late September it was about 1 in 43.

Equally, Wikipedia dropped from 1 in 7 solutions to 1 in 35 over the identical interval, and YouTube from about 1 in 370 solutions to 1 in 500.

These websites are nonetheless amongst ChatGPT’s most-cited domains, however their share has shrunk quick, so if quite a lot of your AI visibility comes by means of them, control how usually ChatGPT cites them to your matters.

The citations Reddit, Wikipedia and YouTube misplaced went to analysis our bodies, regulators, manufacturers and hands-on testers, whereas explainer and aggregator websites fell.

The chart beneath reveals a number of the largest rank adjustments amongst ChatGPT’s 200 most-cited domains, throughout 2.4 million solutions in March and three.8 million in September.

PubMed climbed from sixty fourth to 4th amongst ChatGPT’s most-cited domains and the FDA entered at eighth, whereas well being websites that summarize that analysis, like WebMD and Verywell Well being, fell.

To ensure this wasn’t only a results of new prompts being added to Model Radar, we in contrast ~25,000 prompts that appeared among the many 150,000 highest-volume prompts in each March and September.

Each website within the chart moved in the identical course on these equivalent prompts.

The prompts in any set will have an effect on which websites present up, however the sample is constant: ChatGPT appears to be prioritizing authentic sources over websites that summarize them.

In case your content material primarily summarizes info that another person publishes first, ChatGPT could also be extra more likely to cite the unique supply as an alternative.

The pages holding up finest are those solely you possibly can publish, like authentic analysis, hands-on testing, and your personal product details.

Analyzing fan-out queries in bulk reveals you what ChatGPT checks earlier than it recommends a model in your trade, so you understand what content material to create and the place to earn protection.

A single fan-out question tells you little or no, so it helps to see many facet by facet to acknowledge a sample.

In Model Radar, you possibly can research fan-out queries throughout ChatGPT and Perplexity.

Simply head to the AI responses report, add a filter, and enter your subject as immediate or response textual content.

We’ve chosen “trainers”.

Then filter the platform to ChatGPT and the placement to your market, and have a look at the “Fan-out queries” column subsequent to every immediate.

To check them in bulk, filter the report on the fan-out textual content, corresponding to website:, “official”, “value”, “overview”, or “finest”. You possibly can then research them in-platform or export to a spreadsheet.

Once we analyzed “trainers”, Model Radar returned 1,474 US ChatGPT responses with fan-out queries.

“Finest” appeared 206 instances, “2026” 160 instances, “overview” 117 instances, “vs” 105 instances and “official” solely 65 instances.

So for trainers, ChatGPT principally hunts for current, third-party critiques and head-to-head comparisons, which is a unique image from the general knowledge above.

A working shoe model would get extra from overview protection and comparability pages than from its product pages alone.

That is based mostly on the present prompts within the Model Radar database, however you too can run this evaluation by yourself customized “Tracked prompts”.

Our knowledge agrees with different analysis on the course ChatGPT is heading: towards official sources, website: searches, and fewer citations.

The primary distinction is scale, since different research discovered much more fan-outs and website: searches per immediate.

The prompts in our research are broad.

They stem from actual search demand throughout each subject and nation, whereas many different research used business “finest X” prompts that naturally set off model lookups.

We additionally tracked free, logged-out classes, which search lower than paid or reasoning fashions.

These numbers are a snapshot of 1 mannequin between August and September 2026, and OpenAI adjustments how ChatGPT retrieves info continuously.

When you monitor fan-outs often, you possibly can spot adjustments as they occur, somewhat than months later.

ChatGPT’s free mannequin solutions many questions from what it already is aware of.

And, when it does search, it more and more goes to whoever owns the data, just like the model that units the value.

That leaves two methods into its solutions: be properly sufficient recognized that ChatGPT would not have to look you up, or be the unique supply it finds when it does.

Pages that solely summarize another person’s work are at present those shedding citations, so the pages price investing in are those solely you possibly can publish, like your product details or authentic analysis.

However, keep in mind, this might all shift once more with the following mannequin replace, so maintain a watchful eye by yourself presence in fan-out queries and AI responses!