A SISTRIX analysis of 3.8 million German-language ChatGPT responses found that citation patterns changed after ChatGPT began implementing GPT-5.5.
The analysis, published by SISTRIX founder Johannes Beus, tracked 38 daily samples of 100,000 ChatGPT responses. SISTRIX compared citations in the four days before the change with the four days after it.
The company describes the event as a “ChatGPT Core Update,” comparing it to the way Google Core Updates change search rankings. SISTRIX also notes that the data shows correlation, not proof that the model change caused the citation movement.
For transparency, SISTRIX sells Prompt Tracking, the product used to gather the data for the analysis.
What The Data Shows
On a typical day during the tracking period, citation patterns varied by 1-2%. During the May 22-23 model-identifier change, that number jumped to 47%.
The average number of cited sources per response also dropped, from 30.9 before the change to 28.4 after.
SISTRIX’s comparison of 800,000 responses across the pre- and post-change windows shows a pattern in the winners and losers.
Winners
Reddit had the largest absolute gain, adding 7,007 citations per 10,000 responses for a 59% increase. Reddit was already the most-cited domain before the change.
German mainstream publishers gained across the board. Welt.de rose by 99%, faz.net by 124%, and bild.de by 83%. German streaming and sports platforms also gained. Dazn.com rose 383%, sky.de 157%, and kicker.de 357%.
Specialized tools and knowledge sources gained too. Mapbox.com and openstreetmap.org each rose 83%, and justwatch.com increased 624%.
Losers
International aggregator platforms saw the steepest declines. Indeed dropped 47%, Tripadvisor 53%, and Glassdoor 37-52% across country versions. Expedia and Rome2rio each fell 60%.
Global tech platforms also lost ground. YouTube fell 18%, Wikipedia 14%, and Google.com 22%. Facebook and LinkedIn each dropped by around 20%.
German career and comparison portals list visibility as well, Kununu dropped 46%.
The Industry Pattern
The analysis finds three winning categories in German data: German media, Reddit, and specialized tools all gained. Conversely, three losing categories included: international aggregators, global tech platforms, and German career portals.
Beus interprets the pattern as localization. For German-language queries, ChatGPT responses contained more originally German sources after the change, including established publishers and German service brands.
Whether this was the goal of the model can’t be determined from the data, according to SISTRIX.
Reddit is the exception. Despite being predominantly English-language, it continued gaining. This suggests user-generated discussions across all topics continue to be seen as highly relevant.
Why This Matters
The findings add to growing evidence that ChatGPT citation behavior changes with each model. SEJ has covered Resoneo’s data showing that the GPT-5.3 transition reduced cited domains per response. Separate analysis showed that ChatGPT’s default and premium models already produce different citation patterns.
The SISTRIX data now shows how citation patterns changed around a visible model-version change in a large sample.
Looking Ahead
The analysis covers the German market only. SISTRIX notes that similar patterns have been observed in other markets but hasn’t published that data. The company says it will continue tracking citation patterns through its Prompt Tracking product.
Featured Image: Hypostasis/Shutterstock
