Ahrefs Blog Tracks Stability Of AI Citations
Ahrefs Blog analyzed 37,500 AI citations across five tools, finding that citation volatility is normal and may change with model updates. This affects how AI visibility is measured.
Original reporting · ahrefs.com
Original: “How Stable Are AI Citations? 10 Things We Learned Tracking 37,500 Citations”
Read “How Stable Are AI Citations? 10 Things We Learned Tracking 37,500 Citations” at Ahrefs Blog →
How this was written. Clicktag wrote this article with a language model from Ahrefs Blog’s reporting (model: @cf/meta/llama-3.3-70b-instruct-fp8-fast); it describes that reporting rather than adding to it. The original, linked above, is the full story.
Key points from Ahrefs Blog’s report
- Citation volatility is a normal trait of AI answers and may change with model updates.
- AI citations may not be a reliable KPI due to their instability, and brand mentions could be a better alternative.
- The longer a citation has lasted, the more likely it is to return after disappearing.
Ahrefs Blog recently conducted a study on the stability of AI citations, tracking 37,500 citations across five AI tools over 20 days. The study aimed to understand what happens to citations that drop out and how they can be used as a key performance indicator (KPI).
The study found that citation volatility is a built-in trait of AI answers and may change significantly with model updates. As a result, AI citations may not be a reliable KPI, and brand mentions could be a better alternative since they are more stable.
The research showed that 50-83% of citations disappear by the next day, but about 75% of those that disappear return within 20 days. This suggests that judging AI visibility from one check is not accurate and that tracking citations over a longer period is necessary.
The study also found that the longer a citation has lasted, the more likely it is to return after disappearing. Citations that lasted 8-14 days were more likely to return than those that lasted only one day. This is similar to the Lindy effect, where the longer something has survived, the better its chances of continuing to survive.
To measure the stability of AI citations, the study tracked the number of days a citation appeared over a 20-day period. The results showed that counting how many days out of the last seven a citation appeared is a better indicator of AI visibility than checking only once.
The research also highlighted the importance of separating questions about a brand from broader discovery questions. Being cited for a brand's pricing is different from appearing for a 'best X' question, and tracking these separately can provide a more accurate picture of AI visibility.
The study's findings have implications for how AI visibility is measured and tracked. Rather than focusing on individual citations, it may be more effective to track sustained improvement in citation frequency over time.
Questions and answers
What happens to AI citations that disappear?
About 75% of lost AI citations return within 20 days, with around half returning within three days.
How can AI visibility be accurately measured?
Tracking citations over a longer period, such as counting how many days out of the last seven a citation appeared, can provide a more accurate picture of AI visibility.
Why are brand mentions a better KPI than AI citations?
Brand mentions are more stable than AI citations, which can change significantly with model updates, making them a more reliable indicator of AI visibility.
