Two AI search engines answering the same question share between 4% and 13% of their sources
There is no single "list of sites you need to be on". There is one per engine.
Across 9,124 pairs of answers to the same question, the overlap of cited domains between two engines ranges from 3.9% to 13.4%. In 6 out of 10 ChatGPT–Perplexity comparisons there is not one single source in common.
- Sample
- 9,124 pairs of answers to the same question, compared across engines
- Period
- 26 June – 1 September 2026
- Method
- For each question, week and brand we compare the domains cited by each pair of engines (Jaccard index). Only pairs where both engines cited at least one source are included.
- Language
- Spanish (Spain)
The numbers
| Pair of engines | Answer pairs compared | Average shared sources | Pairs with not one source in common |
|---|---|---|---|
| Google AI Mode ↔ Perplexity | 5,247 | 13.4 % | 11.2 % |
| ChatGPT ↔ Perplexity | 2,004 | 3.9 % | 60.3 % |
| Google AI Mode ↔ ChatGPT | 1,873 | 5.4 % | 68.7 % |
The measure is the Jaccard index over cited domains: how many both engines share divided by how many they cite between them. By domain rather than exact URL because what gets decided with this is where to publish, and that is decided at site level.
Only pairs where both engines cited something are included. Comparing an engine that cites with one that answered without searching would give 0% overlap and would be measuring something else: whether it searched, not whether they agree.
What it means for a content strategy
A site that cites you in one engine does not cite you in the other — and not because it rejected you: because the other engine does not read it.
- Measuring one engine and extrapolating is a methodological error. The list of domains that decides your visibility in Perplexity barely resembles ChatGPT's.
- Appearing on the wrong site is wasted work even when the site is a good one. The useful question is not "is this a reference publication?" but "does the engine my customers use cite it?".
- The closest pair is Google AI Mode and Perplexity (13.4%), and even there eight out of ten sources are different. No two engines are interchangeable.
The extreme case
Between ChatGPT and Perplexity, in 60.3% of comparisons not a single domain matches. Two tools answering the same question in the same sector, in the same week, leaning on bibliographies that never touch.
What this finding does NOT prove
- It does not say which engine cites better sources. It measures disagreement, not quality.
- The ChatGPT figures come only from runs with web search enabled, which are the only ones that cite. They do not describe ChatGPT answering from memory.
- Comparison is by domain, not by URL. Two engines citing different pages of the same publication count here as a match.
- The corpus is Spanish and covers our clients’ sectors. A niche with very few available sources would show higher overlap purely for lack of alternatives.
Frequently asked questions
Which sites do you need to appear on for AI to cite you?
It depends on the engine and the sector, and that is the conclusion rather than an evasion: across 9,124 comparisons, two engines answering the same question share between 4% and 13% of their sources. The useful list comes from measuring which domains each engine cites on your market’s questions.
Can I measure one engine and assume the rest?
No. That is the methodological error these data rule out: in the closest pair, eight out of ten sources are different, and between ChatGPT and Perplexity six out of ten comparisons share no domain at all.
Why compare domains rather than specific URLs?
Because the decision made with this data is where to publish or where to earn a mention, and that is decided at site level. Comparing exact URLs would give even lower and less actionable overlap.