Semrush says 62 percent of citations in AI answers name no brand

Getting named in an AI answer now means asking a stranger to edit their article. Semrush counts that 62 percent of the citations it tracks name no brand at all, and on 21 September 2026 it published a workflow that finds those pages for the questions a team already tracks, marks the ones where the brand is missing or described wrongly, and produces a list of writers to ask for a correction. The post is by Carlos Silva with contributions from Faizan Ali, and the analysis runs in Claude Code.

What the command gives back

The input is a CSV export from Semrush Prompt Tracking, the paid tool that records which pages AI answers cite for the questions you give it. Running /analyze-sources in a Claude Code folder reads each cited page and compares it against a file of approved facts about the brand. It writes three files: pages worth contacting, pages a person has to judge, and pages routed elsewhere. Each page carries one label from a fixed list of six: Not mentioned, Underrepresented, Outdated or inaccurate, Negative framing, No clear opportunity, or Manual review required. In the one worked example, an H&M export of 452 cited pages came down to 330 after duplicates and exclusions.

What you send the writer

Semrush attributes the pitch structure to the outreach specialist Sean Markey: open with one correction, thank the writer for the existing mention, offer the updated information, ask permission before sending a longer list, and follow up after three or four days. Markey says 8 of his last 14 pitches sent this way came back agreeing to take the changes. That is his own recent count over 14 messages, not a response rate anyone should plan against.

The file that decides the result

The judgement sits in brand-facts.md, which the team writes and keeps current: the approved positioning, the pricing, the competitor list and the wrong claims the team already sees repeated. Claude Code only sorts cited pages against that file, so a thin facts file produces a thin list of opportunities. Semrush says a person reviews the output before any message goes out, and the workflow ends at a draft rather than at a sent email.

Who counted the numbers

The 62 percent figure is Semrush's own, from its study with Kevin Indig, which describes the case it counts as "AI might cite your website in a response, but never mention you". Indig is the named co-author of that study rather than an independent check on it, and the 452 page example comes from Semrush's own tool. The workflow also needs a paid Prompt Tracking seat for the input, so a team without one gets the method and has to find its own list of cited pages.

Source

Carlos Silva with Faizan Ali, Semrush, AI citation outreach: a Semrush & Claude workflow, 21 September 2026. The 62 percent figure is from Semrush and Kevin Indig, The Ghost Citations Study.