Which Publications Get Cited by ChatGPT and Perplexity? Six Ways to Find Out
The publications that AI search engines cite are not the same publications that rank first in Google, and they are not the ones at the top of a traditional media list either. Ahrefs found that only 37.9% of the 4 million URLs cited in Google AI Overviews ranked in the top 10 for the query, and 31% did not rank in the top 100 at all. Muck Rack's Generative Pulse found that 84% of the 25 million links cited by ChatGPT, Claude, and Gemini were earned media. If you want a quote or a byline to show up when someone asks an AI assistant a question in your category, you need to know which sources the engines are already pulling from. Here are six ways to find out, and what each can and cannot tell you.
Method 1: Ask the engines directly
Open ChatGPT with search enabled, Perplexity, Google AI Mode, and Claude, and ask the five to ten questions a buyer in your category would ask. Write down every source each engine cites. Repeat over a week, because answers vary by session and by day.
What you get: the publications shaping answers for your exact prompts, with no tool in between. What you miss: scale. A handful of runs can mislead you about how often a source appears.
Method 2: Run an AI citation audit
Several tools automate Method 1. Otterly, Peec AI, and Profound run a prompt set across the engines and return the sources cited, with trials or free tiers for a limited number of prompts.
What you get: a ranked list of cited domains in minutes, and the competitors that appear in your place.
What you miss: history. Citation patterns shift. Semrush's study of 230,000 prompts tracked ChatGPT's Reddit citations falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.
Method 3: Use an AI visibility platform
Semrush's AI Toolkit and Ahrefs Brand Radar track which domains AI Overviews and chat engines cite for keyword sets you define. Profound, Peec AI, and Otterly do the same with prompts rather than keywords, and most show share of voice against named competitors.
What you get: a time series, so you can see whether a placement moved anything.
What you miss: each platform runs its own prompt set, so two tools give two different top-ten lists. Treat the overlap as the signal.
Method 4: Read the public citation studies
Goodie's news publishing study breaks 31 million citations down by publisher and engine, and finds that the top five publishers take 66% of news citations, with Forbes alone at 33%. Muck Rack's Generative Pulse is updated quarterly. Both are free.
What you get: a sourced baseline for the publications that dominate AI citations in general.
What you miss: your category. News and general-interest outlets dominate these indexes.
Method 5: Check whether the publication is ready for agents
A publication that blocks AI agents cannot be cited by the engines that honor the block. Goodie found that The New York Times, which blocks all 13 major AI user agents, received zero citations from ChatGPT and Gemini, though the block did not stop Grok, AI Overviews, or Perplexity. BuzzStream found 79% of the top 50 US and UK news sites block at least one AI training bot.
Free scanners score any domain. Is Agentic, from Vercel with scans run by Ora, scores a site from 0 to 100 against the open AgentReady standard. Cloudflare's Agent Readiness score checks robots.txt, Markdown availability, bot access rules, and API discoverability. You can also read a publication's robots.txt yourself and look for Disallow rules under GPTBot, ClaudeBot, PerplexityBot, and Google-Extended.
What you get: a quick read on whether a placement can be reached by agents at all.
What you miss: readiness is not citations. Forbes scored 45% on Goodie's readiness audit and still holds a third of news citations.
Method 6: Look at readiness where you already pick publications
Some journalist request platforms now show these metrics inline. Connectively displays an Agentic Score from Is Agentic beside Domain Authority and Domain Rating for each publication posting a question, refreshed every 15 days, and reports that across 906 publications the correlation between Domain Authority and Agentic Score is 0.08. Check the column headers in whatever tool you use.
What you get: the readiness check inside the workflow where you decide which requests to answer.
What you miss: the same limit as Method 5.
Putting it together
Start with a week of manual prompts or an audit to see which publications engines cite for your category today. Cross-check against a public study to separate category sources from general-interest ones. Then run the finalists through a readiness scanner, so you know which of them agents can reach. The result usually looks nothing like a media tier sheet: trade publications and niche sites next to a few household names.