Why are citations important in AI visibility analysis?
Citations form a very important part of AI visibility analysis. In the AI first world, the trend is now of zero click searches. This means that you get a ready made answer to your query with careful consideration of the context and intent, without having to click on a URL and go through the contents. With this user flow, how does a person land on a brand or third party website? LLMs while giving an answer also cite the links they have referred to. For marketers this means that optimise your website and digital footprint so that LLMs use your URL and content as evidence and cite the links along in the answers.
Observations in Citation Data in AI Visibility Result Metrics: LLMs are citing old non existent pages:
Each LLM has different models which are trained on different data based on the year of their release. Also, LLMs based on the platforms and user’s paid or free subscription may or may not perform real time searches on the web to get answers. They rely on the existing knowledge they have.
Depending upon the LLM and its model version, information can come from old active indexed pages, cached content, third party datasets, historical versions of the pages and model’s pre existing knowledge.
The key takeaway 1:
A Citation in the LLM answer does not necessarily mean that LLM accessed current live page while synthesising an answer.
The key takeaway 2:
While signing up for an AI visibility or GEO tracking service, make sure the AI models you opt for are trained on recent datasets.
URL may be live while the product or service has been discontinued:
Citations are a key to understand how your digital footprint is performing within LLMs. Not every business is pro active in updating their website with latest information. It may also happen that a product or service has been discontinued but the business still wants to showcase the work they did establishing successful execution over years, hence the URLs may be live. LLMs may or may not get to establish if the information is still relevant or not. This can lead to wrong information being used to synthesise an answer, potentially leading to wrong data or hallucination.
Pro Tip: URLs cited must be categorised based on their relevance:
Businesses analysing citations must categorise them in three categories based on their relevance:
Live Citations - Active URL with revenant information
Historical Citations - URL is cited but the pages does not exist / 404
Hallucinated / Invalid Citations - URL cited does not support the claim
This means that even if a citation exists, The URL can be live or not. If the URL is live, it may or may not support the claim in the answer.
Categorising citations help brands plan their content strategy by identifying gaps where there URLs are not being cited. If they come across any pages with wrong information, they can carry out campaigns to correct them as well. It also helps them to chase URLs and bring them offline which are either historical or lead to hallucinations.