Skip to main content

Instagram /v1/metadata returns an empty description for posts that have a caption (degraded cached response + "ad" posts)

Hi team,

We use GET /v1/metadata to read Instagram captions, and some public posts come back with "description": "" even though the post has a caption. The caption is visible on the post and present in Instagram's public embed data (edge_media_to_caption). We see two separate cases.

Case 1: intermittent degraded response, then served from cache

URL: https://www.instagram.com/reel/DfhrKvKI30i/

  • 2026-11-10 12:50 UTC: HTTP 200, 1 credit, "description": "", "tags": [].

  • For the next ~10 minutes, the same URL returned the same empty response, including with a trailing-slash variant and with "?utm_source=ig_web_copy_link". This looks like the documented ~10-minute response cache.

  • In the same window, https://instagram.com/reel/DfhrKvKI30i/ (no www) and https://www.instagram.com/p/DfhrKvKI30i/ returned the full caption (427 chars).

  • 2026-11-10 15:28 UTC: the original URL now returns the full caption.

The empty response has a different shape from the normal one, so it seems to come from a different extraction path:

degraded: description "", tags [], stats.views 153408, additionalData without "hasAudio" normal: description filled, tags filled, stats.views null, additionalData with "hasAudio"

Every complete Instagram response we have looks like the "normal" shape. The only empty one looks like the "degraded" shape.

Case 2: "ad" posts always return an empty description

URL: https://www.instagram.com/p/DfMth0AAOda/

  • Empty description on 2026-11-06 and again on 2026-11-10, with the /p/, /reel/ and no-www URL forms, including after the cache window.

  • The response has the "normal" shape (hasAudio present, views null), but description is "" and tags is [].

  • Instagram's embed data for this post has a non-empty edge_media_to_caption and product_type "ad". The other posts we checked are product_type "clips". So captions on sponsored/ad posts seem not to be parsed.

Questions

  1. Can you fix caption extraction for both cases?

  2. Until then, can a response flag an incomplete result (e.g. a field saying the caption could not be read), so we can tell "no caption" from "caption not retrieved"?

  3. Is there a way to bypass the cache (e.g. a refresh/noCache parameter), so a degraded response isn't served again for ~10 minutes?

This may be related to https://feedback.supadata.ai/p/intermittent-partial-response-from-v1metadata-for-instagram-reels-author-present (also intermittent partial /v1/metadata responses on reels, but about thumbnailUrl/avatarUrl rather than the caption).

Thanks!

Log in to comment and vote

No comments yet

Be the first to share your thoughts.