Blog SEO

The 100 most cited websites by ChatGPT, Claude, Gemini and Perplexity (2027 edition)

BLOGGING TITAN RESEARCH

By the Blogging Titan Research Team | Published 6 November 2026 | Data collected 2 September and 6 November 2026 | Sample: 100 blogging questions x 4 AI engines | Download the data: All 1,704 citations (CSV) · The 200 questions (CSV)

The short answer. We asked ChatGPT, Claude, Gemini and Perplexity the same 100 questions about blogging, SEO, platforms and making money. The 400 answers linked to 1,704 sources on 906 different websites. Reddit, HubSpot, Ahrefs, Google and Semrush were cited most often overall. The engines share a small core and diverge in a long tail: 68% of ChatGPT’s citations went to sites another engine also cited, but only 4 of the 100 sites ChatGPT cited were cited by all four (Ahrefs, Bluehost, Productive Blogging and TechRadar), and 717 of the 906 sites appeared in one answer only. Perplexity cited sources on every answer; ChatGPT on 52, Gemini on 62 and Claude on 49. ChatGPT and Perplexity were asked in September 2026, Claude and Gemini in November.

Key findings

ChatGPT and Perplexity were asked on 2 September 2026 and Claude and Gemini on 6 November 2026, so engine-to-engine differences below may partly reflect the two-month gap.

• A shared core, a long tail. 110 websites were cited by two or more engines, and they carry much of each engine’s citations: 68% of ChatGPT’s, 45% of Gemini’s, 41% of Perplexity’s and 26% of Claude’s. Only 4 of the 100 websites ChatGPT cited were cited by all four engines.
• Pairs overlap unevenly. As a share of the smaller engine’s sites: Perplexity and Gemini shared 61 of Gemini’s 172 (35%), ChatGPT and Perplexity 31 of ChatGPT’s 100 (31%, the only pair asked on the same day), ChatGPT and Claude 13 of 100 (13%), and Claude and Gemini 15 of Gemini’s 172 (9%), the least.
• The tail is wide. 717 of 906 websites (79%) appear in a single answer. Each question was asked once per engine, so this measures breadth, not stability over time.
• Nearly half the answers outside Perplexity cite nothing (137 of 300). ChatGPT cited no source on 48 of 100 questions, Claude on 51 and Gemini on 38. Perplexity cited sources on all 100.
• The craft-question silence. On the 8 questions about writing craft (headlines, writer’s block, voice), ChatGPT, Claude and Gemini cited a source on 2 of 24 answers. On the 15 questions about tools, they cited on 44 of 45.
• Claude rarely cites on SEO questions. Claude linked a source on 3 of 20 SEO questions, against 11 for ChatGPT and 9 for Gemini.
• Reddit is a Perplexity habit. Perplexity cited Reddit 46 times. Gemini cited it twice; ChatGPT and Claude never did.
• Claude cites free-hosted sites the others ignore. 16 of Claude’s 264 citations went to sites hosted on free developer platforms (pages.dev, vercel.app, netlify.app, github.io) and 10 to dev.to posts. The other three engines together cited free developer hosting once and dev.to never.
• Vendors are the largest group in the top 40. By our classification of each site’s main business, 18 of the 40 most cited sites sell software or services, 2 more are publishers owned by such companies, and 2 are agencies. Four are independent bloggers.
• Blogging Titan was cited 0 times in 1,704 citations. We report that because it is true.

Why we ran this study

The large AI citation studies published in 2026 rank websites across every topic at once, for example Peec AI’s analysis of 30 million cited sources across five AI surfaces (Peec, March 2026) and the AI Platform Citation Source Index, which combines six earlier studies into one ranking (Everything-PR, 2026). They are useful, but a blogger cannot see from them which sites win blogging questions, and neither publishes a question-by-question dataset.

We did the narrower thing and published everything. Same 100 questions, four engines, every citation in a CSV, every question in the dataset. Anyone can re-ask the questions and check the result. This is the second wave of our study: in September 2026 we ran the same questions on ChatGPT and Perplexity (our AI Citation League Table). In November we added Claude and Gemini.

The 100 most cited websites

Ranked by reach: the number of answers that cite the website at least once (out of 400 answers, 263 of which cite anything; 62% of all reach comes from Perplexity). Ties are broken by how many engines cite it, then by total citations, then alphabetically. Ranks 87 to 100 are 14 of the 103 websites tied on a reach of 2; the full list of 906 is in the dataset.

Blogging Titan Research. Source: 400 answers to 100 blogging questions. ChatGPT and Perplexity collected 2 September 2026; Claude and Gemini collected 6 November 2026. Observed. Full data: CSV.

Read the table with one caution. Perplexity lists about 10 sources on every answer, while the others average 3 to 5 when they cite at all, so most of the combined reach comes from Perplexity. The per-engine columns let you rank by any single engine.

What each engine cites most

Observed.

Top 10 websites per engine, by reach (answers citing the site):

Ties listed alphabetically. Observed.

The four lists describe four different habits:

• ChatGPT leans on official documentation and the largest SEO vendors: Google’s own pages, Ahrefs, WordPress.org, Semrush.
• Perplexity leans on Reddit, the big marketing-software blogs and a few long-running blogging sites.
• Claude cites a long tail of small sites, including independent bloggers (Lilach Bullock, Alejandro Rioja), news about blogging (Blog Herald) and pages on free developer hosting.
• Gemini spreads its citations most evenly: its most cited site appears in just 4 of its answers (Gemini’s counts are a floor; see method).

When the engines cite at all

Answers citing at least one source. Observed.

Two topics agree across the three engines that choose when to search: tool questions almost always get citations (44 of 45 answers) and writing questions almost never do (2 of 24). The rest vary by engine. On SEO, ChatGPT cited on 11 of 20 and Claude on 3; on making money, Gemini cited on 14 of 15 and ChatGPT on 7. One reading, which this study did not test, is that engines search when a question has a current, checkable answer (which tool, what it costs) and answer craft questions from what the model already knows.

Who gets cited: vendors, publishers and bloggers

We classified the 40 most cited sites by their main business. This is our judgement, and the classification is in the dataset.

The classification is per site, not per answer: a vendor’s blog can be cited on a question that is not about its own product.

What this means for bloggers

None of this shows that a tactic causes citations. These are the patterns in what the engines chose to link on the dates shown.

1. Check each engine separately. The engines share a core of big sites but diverge in the tail where most blogs sit. Find out which engine your readers use.
2. Topics with current answers draw more citations. Tool questions drew citations from all four engines; writing-craft questions drew them almost only from Perplexity.
3. The tail is wide. 79% of cited sites appear in one answer. In a separate study of Google AI Overviews, our 87-day Citation Lottery re-pull found 34.9% of citations survived a quarter. Check your own citations more than once before acting on them.
4. Reddit mattered for Perplexity, not the others. In this sample, Perplexity cited Reddit 46 times, Gemini twice, and ChatGPT and Claude never.

Method

Questions. 100 questions a new or growing blogger asks, written by Blogging Titan for the September 2026 league table, across eight topics: getting started (15), platforms and hosting (15), tools (15), SEO (20), making money (15), writing (8), AI (7) and admin (5). The full list is in the dataset. Using the same questions lets the two waves be compared directly. Seven questions mention “2026” (questions 1, 4, 14, 16, 29, 34 and 66); they were kept word for word so the waves stay comparable.

Engines and collection.

Questions were typed into each product’s normal chat box at roughly one question a minute per browser tab, the same way a person would ask them. No API was used.

What counts as a citation. A URL the answer links as a source: ChatGPT’s source pills (recovered in full from the page, including sources hidden behind “+N”), Perplexity’s source list, Claude’s inline citation links, and Gemini’s source chips. Each URL is counted once per answer. Pages an engine says it searched but does not link are not counted. ChatGPT’s links inside its prose were excluded in wave one and stay excluded here.

Gemini source chips. Gemini shows sources as chips in the text. A chip can hold more than one source (“WPForms + 1”). Our harvest opened each chip and recorded the source it displayed first, so Gemini’s citation counts are a floor. From question 18 onward (and on question 9, which was re-asked) we also recorded each chip’s label. Those labels show at least 55 more sources hidden behind “+N”, so Gemini’s true count is at least 252, not 197.

Domains. Each URL is reduced to its registrable domain (blog.hubspot.com becomes hubspot.com; .co.uk and similar suffixes are kept). Sites on shared free hosting (blogspot.com, pages.dev, vercel.app, netlify.app, github.io, substack.com and similar) are counted by their own subdomain, because each is a different publisher. Medium and WordPress.com are kept as one site each.

Ranking. Domains are ranked by reach: the number of answers, across all four engines, that cite the domain at least once. Ties are broken by the number of engines citing the domain, then by total citations, then alphabetically. Reach is used instead of raw citation counts because one answer can cite several pages from the same site, and because Perplexity lists about ten sources per answer, more than any other engine.

Checks. Every answer was checked automatically before it was stored: the chat had to contain exactly one question, matching the question sent, and the answer had to have finished. Answers that failed were re-asked up to twice. An independent reviewer then recomputed every figure in this article from the raw files; all counts reproduced.

AI answers vary by account, location, time and wording. These results describe what we observed on the dates shown, not what every user will see.

Limitations

1. Two collection dates. ChatGPT and Perplexity were read in September, Claude and Gemini in November. Some of the difference between engines may be time, not engine. Our own Citation Lottery re-pull found 34.9% of Google AI Overview citations survived 87 days, so citation sets do move. Wave four (re-running ChatGPT and Perplexity) will measure this.
2. One run per question. Each question was asked once per engine. Re-asking can return different sources.
3. Accounts. Claude was run on a paid account in incognito mode, Gemini on a logged-in personal account, ChatGPT on a free account and Perplexity logged out. Paid tiers and other settings can cite differently.
4. Short answers. A few answers are much shorter than each engine’s median (for example Gemini’s answer to the ConvertKit vs Mailchimp question, 934 characters, no citations, the only tools question where an engine cited nothing). We kept them as captured and list them in the dataset README.
5. UK connection. Results may differ elsewhere.
6. Gemini is a floor. Chips holding more than one source were recorded by their first source only.
7. Our questions. The 100 questions were written by us. A different question set would produce a different list.
8. Not a measure of quality. A domain’s reach shows how often an engine chose to link it, not whether its advice is right.

What we will measure next

• Wave three (planned for late November 2026): 100 new questions on affiliate marketing, hosting, content marketing and publishing (already written and published in the dataset), asked on all four engines.
• Wave four: re-ask the original 100 questions on ChatGPT and Perplexity to separate change over time from differences between engines.
• Project 5: the page-level features (bylines, dates, structure, original data) of cited pages against pages that rank but are not cited.

Data and how to cite it

Blogging Titan Research Team. “The 100 most cited websites by ChatGPT, Claude, Gemini and Perplexity (2027 edition).” Blogging Titan, 6 November 2026. Data collected 2 September and 6 November 2026. https://bloggingtitan.com/blog-seo/most-cited-websites-ai/. Licence: CC BY 4.0.

Related Blogging Titan research

• The AI Citation League Table (wave one: ChatGPT and Perplexity)
• Who wrote the 951 pages ChatGPT and Perplexity cited
• 224 of the 951 pages AI search cited carry no publication date
• AI citations per answer: Conductor’s benchmarks vs our sample
• The Citation Lottery, 87 days later
• AI Citation Grader

Change log

• 6 November 2026: First published. Waves one and two, 400 answers.
• 6 November 2026: Dataset v2. Two sites on two-part country domains are now recorded as rebrand.com.my and ithome.com.tw (previously com.my and com.tw). One citation each; no figure in this article changes.

Written by

Blogging Titan Team

Blogging Titan is an independent team of bloggers documenting what actually grows a blog in the AI search era. We have been building, ranking, and monetizing WordPress sites since 2017, and every guide on this site is based on strategies and tools we have tested ourselves. Want a second pair of eyes on your blog? Request a free blog audit or start with the 2026 playbook.

Summarize with: ChatGPT Perplexity Claude Grok