AI Training Data is the New Position Zero 

TL;DR 

AI increasingly relies on pre-trained data to answer queries, meaning if your brand isn’t included in AI training datasets, you risk being overlooked. While AI sometimes performs live lookups (grounded searches), over half of its responses rely purely on internal knowledge. If your competitors’ content is included and yours isn’t, they gain a significant advantage. Ensuring your content is part of AI training data is now as crucial as SEO was in the early days of search engines. 

How does AI training data effect AI responses?

As AI becomes the way more and more people are discovering and gaining confidence in your brand, it is crucial that you have an understanding of what it is saying about you. AI overviews are now the new ‘position zero’, meaning that they can completely overtake what people are seeing in the SERPs. Without you appearing successfully in searches, you are at risk of losing traffic to your site, among other issues. AI is trained on two main areas, live lookups (grounded search) and training data. Live look up are far easier to manipulate, but training data can be the key to making sure AI has a correct and extensive understanding of your site.

Let’s tackle this problem by first seeing how AI searches for content.

AI retrieves information in two primary ways: 

Why Does AI Use Both Methods? 

If all AI models relied solely on external searches or ‘grounded data’ it would be far slower and more costly due to API lookup costs. Therefore, it must have a mix of both to be completely optimised for the benefit of its users. You may ask why it does not rely solely on trained data? Well, that would make for very inaccurate suggestions on simple questions that change daily. It needs up-to-date information to be able to successfully serve the queries it gets.
In short, a balance must be struck in order to optimise speed, accuracy and cost.

Query	AI Confidence Score	Search Type
“Who discovered gravity?”	99%	Standard search
“What’s the latest iPhone model?”	40%	Grounded search
“How many legs does a dog have?”	100%	Standard search
“What’s Tesla’s stock price today?”	10%	Grounded search

Why Your Brand Must Be in AI Training Data 

A study by Semrush revealed that 54% of ChatGPT responses rely solely on pre-trained data (standard searches), while only 46% use live lookups (grounded searches). This means that you are at a significant disadvantage if your website is not included in training data.

If your brand’s content isn’t included in that data, AI may: 

If your content is missing from AI training data, your brand could be at a severe disadvantage, particularly in situations where AI doesn’t search the web for answers. 

How to Be Included in AI Training Data 

We are in the very early stages of really understanding AIO (AI optimisation). As such, there are no fast hard and fast methods on exactly how to get into training data. On that note, it is also sensible to be cautious of softwares claiming that they can fix this for you.

The first step is to see if you actually are in training data using Waikay. Just make a report and check your brand overview. From there, if you want AI models to recognize and accurately represent your brand, consider these key actions: 

• Publish High-Quality, Structured Content  
  For seasoned digital marketers, this advice may sound like a broken record, but these two pillars: “Structured” and “High Quality” have a renewed importance in a post AI world. 
  • Ensure your website contains well-structured, factually accurate, and authoritative content (either original or expert level).
  • Use schema markup (structured data) to help AI understand key details about your brand, products, and services. One way to make sure AI correctly understand your content is to use “Content schema markup”, disambiguating your key concepts. InLinks can help for this. 

• Increase Brand Mentions on Authoritative Websites 
  • AI models train on publicly available data, including reputable news sources, blogs, and industry reports. Getting more backlinks from crappy websites won’t help. Having positive reviews from TrustPilot, Capterra or Yelp can move the needle. 
  • Getting your brand mentioned in respected online publications increases the likelihood of inclusion.

• Optimise Wikipedia and Wikidata Entries 
  • Many AI models rely on Wikipedia and Wikidata for factual knowledge. 
  • Ensure your brand has an accurate, well-documented entry on these platforms. If you can’t, try getting some good / non-parasitic backlinks from these platforms.
  • Entity based internal linking structures can help this. Consider checking out InLinks to help with this – especially for larger site who need an automation option.

• Engage in OpenAI and Public Data Contributions 
  • OpenAI and Google use various datasets for training their models. Where applicable, contribute reliable information to these ecosystems. 
  • Look up data sources that use the Resource Description Framework. A list of application is included at https://en.wikipedia.org/wiki/Resource_Description_Framework as these data sets tend to be readily available for AI to use for training.  
  • Some other data sources to consider are: 
    • Common Crawl (https://commoncrawl.org) contains a free, open repository of web crawl data without IP licensing issues
    • Product Hunt (https://www.producthunt.com/) lists new products as they launch, by maker. 
    • IDMB (https://idmb.com)  lists TV, film and Podcast personalities 
    • OpenStreetMap (OSM):  This is a collaborative, open-source map of the world. 
    • Kaggle:  This platform hosts datasets that users can analyse and contribute to
    • GitHub:  While known for code, GitHub also hosts many open data repositories. 
• Monitor AI Model Outputs and Provide Feedback 
  • Regularly test how AI models (e.g., ChatGPT, Google Gemini) present your brand. Use a tool like Waikay that will help you monitor not only your brand, but also your brand related products, services and topics. 
  • Where errors occur, analyse the brand speech to identify missing elements or contradictions. 

• Develop AI-Friendly Content with Clear Branding 
  • AI learns from repeated patterns—make sure your brand’s core messages and key facts are consistently stated across platforms. 
  • Avoid jargon or overly complex phrasing that AI might misinterpret. If you can’t or think that your technical jargon should be included in training data (giving you a potential competitive advantage), then consider building some additional documentation (lexicon, glossaries, “jargonpedia”), both on your site and in external resources. 

(Note: This list is for guidance only. There is no guarantee that being in these data sources will consistently work, and this is a tiny list to get you started.)

In conclusion: it is important to optimise for AI

As AI-driven search grows, being part of AI’s training data is the new Position Zero. Brands that fail to adapt will lose visibility, while those that embrace AI inclusion will dominate in the new search paradigm. 

Written By: Fred Laurent
Edited by: Genie Jones