What AI engines cite when you ask about LLM gateways

What we did

We wrote 27 questions the way a developer asks them, not the way a keyword tool suggests. For example:

What are the best OpenRouter alternatives?
Is it safe to route API calls through a third-party LLM gateway?
Are free LLM APIs reliable enough for real work?
How do I point Claude Code at a different API base URL?

Each question went to four engines on 2026-09-20. Only one of them returns the sources its answer was built from: Perplexity, reached through OpenRouter. Every other engine answered without exposing citations, and an answer with no citation trail cannot be measured, so it is not in this dataset.

What remains: 613 citation entries across the 27 answers, collapsing to 569 unique URLs and 346 unique domains. That is the whole dataset. It is one engine on one day, and the limits section at the bottom is not decoration.

1. No single source dominates

The ten most-cited domains account for 18.6% of all citations. The remaining 81% spreads across 336 domains, and most of those appear exactly once.

This matters because the standard advice in this space (get into the big awesome list and you are done) does not survive the data. There is no list to win. There is a long tail, and each entry in it is a small bet.

2. The head still has a shape

Domain counts are not random. The frequent ones cluster into three kinds, and the kind predicts the format you would need to publish:

The GitHub entries are not documentation. They are lists (directories of free APIs and gateways) and issue threads, where someone asks how to change the base URL in Claude Code or Cline and other people answer. Both are places where a project can be named without writing a blog post.

The dev.to entries are all first-person and all contain numbers. Not one of them is a product announcement.

And competitor marketing pages get cited directly — including, for the question about whether a gateway can read your prompts, a competitor's privacy policy and sub-processor list. Those pages are usually written for lawyers. They are being read as answers.

3. Reddit barely appears

Three citations out of 613 (0.5%) went to reddit.com.

That is worth pausing on, because "Reddit shows up in 40% of AI answers" is repeated everywhere. The figure is real. It is also averaged across every topic imaginable. For developer tooling questions, the engines reached for repositories and tutorials instead. If you were planning to spend this quarter on Reddit, the data says the quarter is better spent elsewhere — for this category, at least.

4. Where YouTube appears, it is someone else's tutorial

All nine YouTube citations are setup tutorials. Eight of them are about one competitor, by name, walking through a configuration the viewer could copy.

What this implies

Limits

One engine. One day. 27 questions, all of them ours, chosen to cover the questions we care about. A different question set or a different date would produce a different distribution. Three of the four engines returned no citations at all, because they do not expose them — which means this study can say nothing about how those three answer, only about which one is measurable. Treat it as a snapshot, not a trend.

Disclosure

We operate FreeModel, and our product is one of the things these 27 questions ask about. In this round it received zero citations. Every competitor named above was cited more often than we were.

We are publishing the method and the raw distribution rather than a conclusion, because the number that matters is the one you can re-run.

How the study was run

A method is only worth publishing if someone else can repeat it and get their own number. This is the whole procedure, in the order it happened. It is also the source for every count on this page.

1. Step 1: write the questions. 27 of them, phrased the way a developer types them into a chat box rather than the way a keyword tool suggests. No question was rewritten after the first answer came back.
2. Step 2: pick one date. All 27 went out on 2026-09-20. Answers are generated live, so a different day is a different dataset and is not comparable to this one.
3. Step 3: keep only what can be measured. Of the four engines, only Perplexity returns the sources an answer was built from. An answer with no citation trail cannot be counted, so the other three are excluded rather than scored.
4. Step 4: count in three passes. Citation entries first, then unique URLs, then unique domains. Reporting all three is what stops one page cited repeatedly from looking like broad coverage.
5. Step 5: publish the distribution, not the conclusion. The table and the counts ship next to the reading, so a disagreement can be checked against the numbers rather than against the summary of them.

Citation study FAQ