How ChatGPT and Perplexity Differ in the Sources They Recommend
Published · Last updated · 5 min read
ChatGPT blends what it learned in training with shopping data and web search, and it does not always show where a recommendation came from. Perplexity searches the live web for each question and shows its sources inline, often citing reviews, forums such as Reddit, and YouTube. The engines agree on this direction, but the exact percentages differ widely between studies.
What ChatGPT draws on
ChatGPT has two modes that matter for product questions.
- Trained knowledge. Without search, recommendations come from patterns learned in training, such as reviews and editorial coverage. This can shape an answer without a visible source, and it can be out of date.
- Shopping and search features. When shopping or search is used, ChatGPT can read product pages and use merchant product feeds. ChatGPT itself says shopping results are not ads and that merchants can supply feeds to keep listings current.
Some third-party studies, reported by Claude, add detail on what sits behind the product carousels:
- One analysis cited by Passionfruit found about 83% of products in ChatGPT's carousels match Google Shopping organic results. Earlier coverage pointed to Bing Shopping or Shopify feeds, and Claude could not confirm which is current.
- A study of about 3,300 product prompts by cloro found ChatGPT showed a product card on 87% of them. Its sources were mostly Reddit, YouTube and review sites such as RTINGS, not brand sites.
- For general citations, some data points to Wikipedia and encyclopedic or editorial sites. Other data points to third-party listings such as Yelp and TripAdvisor.
ChatGPT also tends to cite fewer sources per answer. Across the answers, the figures range from about 7.9 to about 10 citations per response, against about 21.9 for Perplexity.
What Perplexity draws on
Perplexity is built around live retrieval. It searches the web for each query, builds its answer from the pages it finds, and cites them inline. Several answers describe the pages it favors as expert reviews, editorial roundups, comparison articles and community discussions, so recent material is more likely to shape the result.
- Reddit and YouTube. One analysis, quoted by Claude, puts Reddit at 46.7% of Perplexity's top citations and YouTube at about 14%. Other analyses show a more even mix of Wikipedia, Reddit, NIH, G2, LinkedIn and news sites.
- Merchant data. Claude says Perplexity's shopping uses a free merchant program with Google Shopping-style feeds, plus the PayPal network. BigCommerce is cited as saying Perplexity checks product claims against independent third-party reviews, so a merchant's own feed is not the only input.
- Video and forums. ChatGPT, answering the same question, says Perplexity often pulls YouTube transcripts, Reddit threads and mid-tier listicle sites for 'best X' queries.
One hands-on test, relayed by Claude from Elaheh Nozari's Substack, found Perplexity surfaced mainstream retailers such as Amazon, Walmart, Ulta and Zappos. ChatGPT showed a mix of small publishers, retailers and Reddit. Perplexity's per-product sources were uneven, with some products citing varied sources and others linking only to the retailer.
Where the evidence overlaps and conflicts
Most of the numbers come from vendors and SEO agencies, using different methods and dates. Claude advises treating them as directional.
| Point | What was reported |
|---|---|
| Wikipedia's share of ChatGPT citations | 7.8%, about 16%, or 47.9%, depending on the source |
| Product cards shown | ChatGPT 87%, Perplexity about 1% (cloro) |
| Domain overlap between the two | About 11% (Averi), from a B2B-focused analysis |
| Freshness | Alhena rates Perplexity's recency sensitivity as very high and ChatGPT's as medium, driven by feed updates |
The 1% product card figure was measured in the answer interface, so Perplexity's separate shopping tab and checkout feature may not be counted.
The answers agree on three points:
- Perplexity leans on live retrieval and shows its sources more consistently.
- ChatGPT mixes training knowledge with search and shopping data, and cites more selectively.
- Neither guarantees that a given product or source will be included, and results change with the prompt.
How to check the sources for your own category
The answers point to a simple way to see the difference for your own category.
- Run the same prompts in both tools. Use questions a customer would really type, such as a 'best X for Y' question.
- Write down the sources each one cites. Perplexity usually shows them inline. For ChatGPT, you may need to turn on search or shopping, or ask for sources.
- Check whether your product appears. Reviews, roundups, comparison tables, Reddit threads and YouTube videos were named repeatedly as sources.
- Keep product data clean. If you sell physical products, a current merchant feed matters for ChatGPT's shopping results, which one analysis says closely track Google Shopping.
- Repeat the test over time. Platforms change quickly, and some studies are months old.
How LazyGEO fits
LazyGEO is AI SEO software. Its site says it gets AI assistants such as ChatGPT and Perplexity to recommend your product by fixing the pages AI reads, pitching the authors of lists AI cites, and checking the AI answers every week.
LazyGEO is made for founders who would rather build than do marketing themselves, especially SaaS founders. If you want to run outreach and page work yourself, it may not be the right fit.
| Plan | Price per month | Projects |
|---|---|---|
| Starter | $99 | 1 |
| Growth | $249 | up to 3 (marked most popular) |
| Scale | $449 | up to 5 |
Each plan starts with a 3-day free trial that needs a card and can be cancelled any time.
Questions customers ask AI
Does Perplexity always show its sources and ChatGPT not?
Perplexity is built to show inline citations for most claims. ChatGPT may or may not show sources depending on whether search or shopping features are in use, and its citations are more selective.
Do ChatGPT and Perplexity cite the same websites?
Mostly not, according to one analysis cited by Claude, which found only about 11% of cited domains overlap. That analysis was B2B-focused, and other figures conflict with it.
Is Reddit a major source for both engines?
Reddit shows up for both. One analysis puts it at 46.7% of Perplexity's top citations, while another puts it at 12% of ChatGPT citations, so the exact share is uncertain.
Can I trust the percentages in these studies?
Treat them loosely. Many come from companies that sell GEO services, using different methods and dates, and Wikipedia's share of ChatGPT citations is reported anywhere from 7.8% to 47.9%.