Skip to content

Convo with Grok on 2026-09-01

Chrome extensions that use AI and link to Wikidata

It does not really exist in the form you are describing. Pieces of it do.

What you are imagining is: AI reads the page → extracts people, places, orgs, works → disambiguates them to Wikidata QIDs → overlays links / facts / a graph. That stack is named-entity recognition plus entity linking. Libraries can do it. A polished consumer Chrome extension that does all of it on any page is still missing.

Closest things that actually exist

Wikidata, but not AI over the page text

  • Entity Explosion is the best Wikidata browser tool. If the URL matches a known identifier (IMDb, Spotify, YouTube, Wikipedia, iNaturalist, ~5,000 other sites), it resolves the page to a Wikidata item and “explodes” same-as links. Icon lights up when a match is possible. No LLM. It does not scan the article body and link every mention.
  • DataQuest is the same idea as an academic demo: match the current URL to a Wikidata identifier, then show facts and next-hop links. Source: github.com/dpizarrow/dataquest.
  • Wikidata for Web (Q99894727) shows Wikidata on other sites and can extract identifiers back into Wikidata. Editor tool, not an AI overlay.
  • Wwwyzzerdd for Wikidata lets you edit Wikidata from Wikipedia and tag page links with properties.

Highlight a phrase, then look it up (optional AI)

  • Knowledge Tooltip is the closest “AI + Wikidata” combo. Select text → Wikipedia summary, Wikidata facts (born, occupation, population, etc.), and an AI tab if you add an OpenAI key. You have to select; it does not auto-link the whole page. Tiny user count. Source: github.com/ibralassaf/knowledge-tooltip.
  • Wikidata One Click Info / 1-Click-Info: highlight a word → Wikidata label, description, sitelinks. Wikimedia-grant project, lookup only.
  • Right-click “search this on Wikidata” extensions exist; they just open a search tab.
  • Wikilinker is the thing people keep reinventing. It scans the page, matches proper nouns against a local bloom filter of the top 1M Wikipedia titles, and inserts Wikipedia links. Fully local, no API, no Wikidata QID, no disambiguation beyond string match. That is 2004 Wikiproxy logic with a better pack file.

AI, but not Wikidata grounding

  • Wikimedia’s experimental Citation Needed used ChatGPT + Wikipedia to fact-check a selected sentence. Not entity linking.
  • Bookmark/tab graph extensions (ACK, Weft, TabGraph) build a personal graph with an LLM. They do not resolve nodes to Wikidata.
  • Schema App Highlighter does entity linking for schema.org markup. Enterprise SEO, not a Wikidata reader overlay.

Why the obvious product is still missing

Full-page linking is the hard part, not the APIs.

Wikidata search, SPARQL, the new Wikidata embedding / vector DB, and Wikidata MCP already exist for machines. Entity-fishing / spaCy-fishing already link spans to QIDs in Python. Putting that in a Chrome extension means: run NER on messy DOM, disambiguate “Paris” / “Jordan” / “Apple” from page context, not hammer WDQS, stay under Manifest V3 limits, and not spray false links. Wikilinker dodges all of that by only doing title-string match to Wikipedia. Entity Explosion dodges it by only matching the URL. Knowledge Tooltip dodges it by only running when you highlight.

If you want the nearest installable stack today: Entity Explosion for “this page is Q…” plus Knowledge Tooltip or One Click Info for “this phrase is Q…”. The missing product is the one that walks the whole article, writes QIDs onto every mention, and lets you open that as a graph.