← The Brief

Listicle share fell because ChatGPT's source pool grew 75%, Konitzny finds

Indexing absolute weekly volumes against a July baseline separates a dilution effect from a real retrieval drop for Reddit.

Niklas Buschner
Screenshot from David Konitzny's post
Screenshot: David Konitzny

Two readings of the GPT-5.6 switch have circulated widely: listicles down, Reddit down. Both figures came from share-of-source charts. David Konitzny, writing on LinkedIn on 11 September 2026, reindexed the absolute weekly volumes to their July average to check what moved.

Total sources retrieved per answer rose from 100 in July to 175 after the switch, on Konitzny's index. That alone changes every share number downstream.

The listicle drop is a denominator effect

Listicle volume moved from 100 to 92 on the same index. That is practically flat. The share halved because the surrounding pool grew by 75%, not because ChatGPT pulled fewer listicles.

Reddit sits in a different category. Absolute Reddit volume fell from 100 to 31, a two-thirds drop measured against a pool that also got bigger.

Konitzny also traced the May Reddit spike some readers remember. Between 4 and 25 May, ChatGPT appended the word "reddit" to its fan-out queries, pushing Reddit volume to 376 against the July baseline. The peak was a query-construction artefact, not organic demand for the source.

What to check before changing anything

We think the useful move here is to look at your own citation data the same way Konitzny did. Pull absolute counts per source type per week, index them to a stable baseline month, and separate that from share.

A share can fall because a source type is retrieved less, or because everything around it grew. The action is different in each case. Steady listicle volume means retrieval behaviour toward that format has not changed. A real absolute drop, as with Reddit, is worth watching over several more weeks before rewriting anything.