Optimising Product Pages for LLMs / AI Search

We could forgive you for thinking that there is a bit of an 'AI' overload at the minute. It seems to be getting mentioned everywhere. From disrupting social media with sometimes amusing and sometimes concerning content, to the idea of replacing of us all at work.

Every eCommerce event seems to have some thread around AI. Be it automating content creation at scale, through to advertising or data enrichment. In this piece, we review the current state of 'Agentic Search' and look at what you can be doing today to increase visibility of your product pages (PDP's) and future proof your eCommerce web site for tomorrow.

The State of Current Search Trends

The great thing for us in eCommerce is that Large Language Models - LLMs for short, are now quickly becoming part of the everyday tools that our potential (and existing), customers are using. At present it is predicted that between 8 - 9% of web searches are happening within LLMs. This however is growing at an astonishing rate. Popular SEO research platform SEMRush have published a recent research piece suggesting that LLM search will over take traditional search engine searches by 2028.

Alongside this erosion of typical search engine dominance, a recent article from Search Engine Journal suggests that even within the typical search engine result pages, changes are afoot. Research suggests the click through rate of the top search position are down 32% on the previous year. This is mainly because of Google's 'AI Overview' increasing it's impressions on a wider number of search sessions dramatically.

What Does This Mean for Online Shopping?

The key behavioural change here is that AI is being used and trusted to provide a super-fast and highly trusted synopsis of a specific product or brand. 'Agentic Search' is becoming the terminology used when AI agents (think bots), are sourcing and surfacing product information within answers to user queries. They don't just wait for exact queries to be typed either. The system reads your intent, splits complex questions, and runs multiple searches at once in an attempt to provide a rich and aggregated result set gathered from multiple sources in a matter of seconds.

Potential customers are asking queries and questions in a much more conversational manner when researching a purchase via AI.

The instant aggregation of huge amounts of online data also makes it far easier for would be customers to read reviews and gather real world feedback from previous customers via various sources of user generated content. These can include, review sites, chat forums such as Reddit or your own web site as sources.

These are considered to be at present 'Upper MarketingFunnel' type searches at the discovery phase or initial customer intent. However these processes will quickly start to move lower down the funnel, even to the point of conversion happening on AI platforms. Shopify in the U.S. have already announced their partnership with ChatGPT and are allowing direct purchases of a single cart item per transaction within ChatGPT directly via their integration. This is definitely going to become more widespread as integration points grow and more vendors adopt this offering.

So How do I Get My Web Pages & Products Included by LLM's?

The key here is to understand how many AI tools such as ChatGPT, Perplexity & Google's Gemini source their data. The main change from SEO of old is that these platforms don't crawl and index pages as quickly as Google or other search engines might. You also don't have tools such as Google's 'Search Console' or Bing's 'Webmaster Tools' that allow you to add XML Sitemaps and even request indexing of specific URL's.

AI tools have a knowledge base of information behind them. This is a highly organised and structured data set that is populated by training content. This content includes online material. The inclusion is no longer instant and even once your content is included, it needs to be deemed as relevant enough information for it to be cited as a useful data source.

OpenAI have also launched the ability to send your entire catalogue data to the platform via a feed. This is very similar to Google Shopping feeds. The integration makes sure that ChatGPT uses your feed to ensure up to date product detail, stock & pricing data is available in relevant chats.

At the time of writing, we don't have the same update from Google ref Gemini but the logical step here would be that any Merchant Centre feeds could share data with Gemini in the very near future.

How Can I Make Sure my Product Pages are Optimised for LLMs?

The good news here is that some of the previous best practises for SEO are still relevant. But by relevant we mean as a starting point....

The stand out best way to make your content easy for LLMs to read is our old friend 'Structured Data'. (If you want the quick 101 on structured data, see Google's great explanation page here)

Structured Data is a very organised way of listing information about the content of a web page. In this case your product page/PDP and therefore, the product being displayed.

The preferred format at present for this data from many sources is JSON/LD.

This is a separate array of specifically organised information that can be quickly read by any bot crawling a page as it doesn't require the load & processing time needed for the entire HTML to be parsed & processed before the relevant pieces of information can be identified and extracted for indexing. This is also why separate JSON/LD is preferred as an object on the page, over HTML vs inline microdata tagging. As per above, this is slower and requires much more of the pages HTML to be parsed & processed for the data to be extracted. Depending on the eCommerce platform you use, there are several great plugins out there for implementing structured data schema across your site. Alternatively, you could ask your developers or agency partners to implement this directly into the PDP template of your site. Most common product attributes are already available to the page and if not, it's a good exercise on making sure you have as much data as possible available to support agentic searches.

Going beyond Google Supported 'Rich Snippets' for Agentic Search

Another key point to take note of is the rich depth of product spec and information that can be included within valid schema.or markup for the Product type.

Google offers their own testing tools for checking the validity of structured markup on a page. It's important to note though that these only test the attributes that Google uses and supports.

By incorporating many more of the available attributes and types of schema.org, you can make sure you are feeding any agentic search tools with structured and detailed product data that really gets to the deeper questions a consumer may be asking about a product or it's spec.

If your eCommerce platform doesn't support all of the fields that you need to describe your product attributes, consider using custom/meta fields at product level to store information you might not show on the page to users, but can include in any structured data.

Your Quick Wins Product Page Checklist for Optimising Product Pages (PDP's) for LLMs / AI search.

• Review the information you store or present to users at product level. Can you answer the questions that could be asked when comparing to a competitor?

• Make sure that JSON-LD product schema is complete and validated on all product pages.

• Review your product feed setup (Google Merchant Centre, Shopify feed, etc).

• Keep product descriptions factual, not promotional. AI tools prioritise accuracy over anything else.

• Publish customer Q&As and reviews in structured markup for richer LLM context whilst adding genuine authority to the product.

• Monitor referrer data from ChatGPT or Perplexity (where available). make sure you are able to track and report on traffic sources from AI based tools and look for trends changing after making any changes to the above.