Coverage

Coverage is the honest edge of BackTweets: what we can and can’t see. Paste a term or a link and you get three different views of it - the social conversation, the link trail, and how visible it is when AI assistants answer questions. Each one has a different reach, and this page says plainly where each stops.

Social conversations

We search public posts across 20 networks: X, Reddit, LinkedIn, Instagram, Threads, TikTok, YouTube, Bluesky, Stack Overflow, Stack Exchange, Substack, Medium, Hacker News, GitHub Discussions, Product Hunt, Pinterest, Tumblr, Facebook, Mastodon, Quora. Every result is re-checked against our own rules before you see it, and its details are read from the source post itself - so what lands in front of you is an individual public post, not a profile or a feed page.

The limits matter. We cover public posts that are publicly discoverable: never a firehose, never a complete archive, and not real time. Private accounts, deleted posts, and networks that restrict public access are invisible to us, and a brand-new post may take a while to become findable anywhere. Some networks expose far more of their content publicly than others, so a quiet platform in your results may mean a closed network rather than a quiet audience.

Link trail

For a URL or domain we add its history and its links: archived snapshots going back to the page’s earliest capture, a backlink profile (referring domains, total backlinks, domain authority, and the most recently detected links), and public mentions in places like Hacker News, Reddit and Wikipedia. Backlink data comes from a third-party index, which means it is a large sample of the web’s links rather than every link that exists; very large domains can return headline totals without the detailed list. When a profile is unavailable we say so on the report instead of quietly showing you history alone.

AI visibility

We also measure which brands, sites and pages AI assistants mention and cite when they answer questions - reported as a share of AI mentions across a disclosed sample of engines, never a complete census of every AI answer. That sample spans the major AI answer engines, including ChatGPT and Google’s AI Overviews, and starts with US, English-language questions. Think of it as a window, not the whole room: wide enough to show real movement, and disclosed so you always know which engines a figure was measured across. We score domains and sources, individual pages and URLs, and topics; whatever the entity, the number is its slice of AI mentions within the sample.

What assistants have said is only half the picture. A link report also checks whether a site is legible to them at all: content present in the raw HTML, robots.txt rules for AI crawlers, structured data, and machine-readable files. The crawler, structured-data and file checks are provided by Geordy, built by the same team as BackTweets; the raw-HTML check is ours. They are findings rather than a score, they run live against the site each time, and none of them predicts citations.

Coverage grows as more networks, engines, regions and languages come into view; we add to the window deliberately and disclose it when we do. How the sample behind each figure is framed is on Methodology.