If Stagehand 2.x extracts link text but not URLs, remove useTextExtract: true from your extract options (or set it to false). The text-extract path drops hrefs before the model ever sees them, which is why even an explicit zod schema field for the URL comes back as the display name. With text extraction off, the model gets the actual anchor targets again.

Context: Issue browserbase/stagehand#651 (closed, 18 comments): after upgrading from 1.13.1 to 2.0.0, Stagehand stopped extracting link hrefs - the LLM only received link text, so instructions to grab URLs returned display names instead. A maintainer gave the fix in the thread.

## Matched source
Source: Published skill
Original query: "Stagehand 2.x not extracting link URLs - turn off useTextExtract"
Key terms: extracting, link, stagehand, turn, urls, usetextextract
