Update: Scribble Network now tracks Claude
Hey everyone!
First update since the launch, and it's one a bunch of you asked about in the launch thread.
Scribble now tracks Claude alongside ChatGPT, Gemini, Perplexity, Copilot and Grok, so that's six engines, every tracked query, fetched every 24 hours.
Why this matters: every engine we've added so far has had its own sourcing personality (Gemini leans creator content, Grok will build an answer from a single post, GPT leans owned pages).
Early read after the first fetches: Claude picked up our creator content immediately, 14 creator pieces cited across four client brands on day one of tracking. And it has a clear favourite surface so far: almost every one of those citations is a Medium post (one Paragraph piece snuck in).
This is a small sample ofc, and we'll share a proper breakdown once we have enough fetches to trust the pattern.
If you ran an audit during launch week, your tracked queries now include Claude citations automatically. Worth a re-look to see where you stand on it.
More coming. And as always, if Claude surprises you on your own brand, tell me; that's the fun part. :)
Replies
Dial
the Medium bias for Claude citations is a really useful early signal, especially if that holds up as the sample grows. curious whether that's Claude actually preferring Medium's writing style/structure, or just an artifact of Medium being disproportionately well indexed/crawled compared to other creator platforms. those would suggest very different plays for where to publish next.
Scribble Network
@omri_ben_shoham1 Good question, honest answer I can't fully split them yet, but my bet is mostly indexing. Medium's crawled fast by everyone (Claude cited these on day one of tracking), Paragraph's a small platform so under-crawling is the default.
We've built an experiment already by placing near-identical pieces already going out on Medium, Substack and Paragraph at once, so we'll soon have same-content, three-platform data per engine. T
Breakdown coming.
Meridian
Interesting signal. If Claude is already picking up creator content that fast, the Medium concentration feels worth separating from the broader “creator content works” takeaway.
I’d be curious whether you’ll test the same piece published across Medium, Paragraph, Substack, and owned blog formats to see if Claude is responding more to the platform, the structure, or just crawlability.
Scribble Network
@akarsh_hegde That test is already running, near-identical pieces going out on Medium, Substack and Paragraph at once, so we'll have same-content, per-platform data per engine soon. We already create on-site content for the brands we work with before getting creators posting.
And agreed, "creator content works" and "Medium gets crawled by everyone" are two different findings that happen to overlap right now. The breakdown will keep them apart.
@akarsh_hegde @kaavya_prasad The three-platform test controls the content variable, but I think the same confound can still eat it unless you also control for time. If all three go out simultaneously and Medium is crawled fastest, Medium wins every fetch on recency, and you've measured crawl latency a second time rather than preference.
Two things that would separate them. Stagger the arms so each platform gets the head start in one of them, and see whether Medium still wins when Paragraph had a three-day lead. And start counting at confirmed-indexed per platform rather than at publish, then compare citation share over equal exposure windows. Otherwise the piece indexed on day one banks citation-days the others never had a shot at, and the count is partly just a clock.
If Medium still wins on equal exposure, that's a real preference finding, and a much harder one to argue with than the raw count.
Scribble Network
@akarsh_hegde @narek_keshishyan This is a really good catch, and you're right, taking both fixes.
One honest issue on the indexed bit there's no direct way to see inside each engine's index, so we'll have to proxy it. The first appearance in that engine's retrieval, or site, checks on the underlying search layers. Not perfect, but better than counting from publish.
And agreed on the payoff, if Medium still wins on equal exposure, that's a real preference finding. If it doesn't, that's arguably the more useful result, it means the platform gap is just crawl reach, which is fixable. Either way the experiment got better today, thank you
Dial
@kaavya prasad the same-content three-platform experiment is exactly the right way to actually answer this instead of guessing from correlation. if it does turn out to be pure indexing speed rather than genuine model preference, does that change how you'd advise people to publish, e.g. always mirror to Medium day one regardless of where the original home is?
Scribble Network
@omri_ben_shoham1 Agreed, I'll test this out and keep you posted. :)