AI Search / Schema markup question

Squiz Matrix General Discussion
1

Afternoon all,

Quick question on Schema markup for GEO and improving how AI engines use our content to surface answers.

We are about to implemented Squiz’s AI convo search and to prepare for this I am exploring implementing schema markup via a metadata schema, fields and a JSON-LD block in a global paint layout so it appears on all pages.

Is this something anyone else has implemented, any best practice to follow or issues?

JSON-LD block I am going to use:

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "NewsArticle",
  "headline": "%asset_name%",
  "description": "%metadata_description%",
  "datePublished": "%metadata_publish_date%",
  "author": {
    "@type": "Person",
    "name": "%metadata_author%"
  }
}
</script>
2

Hi innesz,

Thanks for sharing this - really useful timing as we’re working through something similar.

Quick question on your AI search implementation: are you planning to surface it more like a traditional search experience (Google), or as a conversational chatbot interface? We’ve set up conversational search on our end (DXP) but haven’t fully landed on the interaction model yet, so would love to hear how you or others are approaching it.

3

Hi @NickyP

Yes, the plan is to implement this to improve the search experience for users. We witnessed citizens, during user research sessions, attempting to ask questions in our search (when we launched our new homepage). This will hopefully solve that problem and reduce recurring enquiries.

We’re planning a hybrid of our current list of keyword search results and the AI result appearing above this - see Greenwich University who launched theirs last year and has seen significant conversations and understandings on our 17-18 years use search and what they are searching for - https://find.gre.ac.uk/s/search.html

If you ask a question on campus or lifestyle as they have only implemented the AI convo search their lifestyle content.

We are targeting a subset of our top 20 recurring enquiries, over a two year period and hope to see a reduction in these by directing users to the new AI search. Kick-off project was yesterday so a lot to do until then.

Please keep in touch on how things progress and I will do the same.