一项 AEO 实验向 ChatGPT、Perplexity、Gemini 和 Claude 提出同一组问题,统计其引用的 1,329 个来源后发现,被引用网站 DR 中位数为 86,仅 7% 的引用来自 DR 30 以下站点。
I ran an AEO experiment and found something critical
the median DR of websites AI cites is 86
I asked ChatGPT, Perplexity, Gemini and Claude the same set of questions and pulled the DR of all 1,329 sources they cited
say you ask ChatGPT “best CRM for startups”. it searches, reads a bunch of pages, cites around 7 sources and writes you an answer
you read the answer, but the AI read the blog posts first and decided who gets quoted
here’s what the cited pages have in common:
- trust
only 7% of citations came from sites under DR 30
AI quotes sites the web already trusts, and that trust is built with backlinks
- page format
73% of the sources were blog content: listicles, articles and comparisons
product and landing pages were 7%, and forums and social were only 3%
- original data
ChatGPT is the exception: only 18% of its citations were blog content, and 63% were the original source, like a pricing page or the docs
AI quotes the place a fact comes from, so if your post repeats the same stats as 40 other posts, there’s nothing in it to quote
- interactivity on the page
many cited pages had a calculator, chart, table or template
the few humans who click through already read the AI’s summary, so they click to do something, not to read
now you can’t get to DR 86 right away, but you can do 2 things:
1. get to DR 30+ in less than 5 months
2. get your product mentioned in the exact listicles AI keeps citing
Outrank helps you do both
it builds backlinks for you automatically and gets your product the #1 spot in the listicles AI cites
so after this experiment, we rebuilt all 280 posts on the Outrank blog this week. 275 of them now have a calculator, chart, checklist or template
that’s what we’re building into every Outrank article: real data, backlinks that raise your DR, and pages worth clicking
so now we write for the AI first, then give the human a reason to click
来源:Tibo · x.com