AI Training Data is the New Position Zero

3 March, 2025

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:

1. Grounded Search (Live Lookup)

• When you ask AI, “What’s the weather like today?”, it doesn’t rely on pre-existing knowledge.
• Instead, it fetches real-time data from the web, ensuring up-to-date responses.
• This method is used for dynamic topics like news, sports scores, or stock prices.

2. Standard Search (Internal Knowledge Retrieval)

• If you ask, “How many legs does a dog have?”, AI pulls from its pre-trained knowledge rather than searching the web.
• This saves time and ensures faster responses.

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.

How AI Decides Between Standard and Grounded Search

AI determines whether to use a standard search (internal knowledge) or a grounded search (web lookup) based on several factors:

1. Query Intent Classification
   • Static/Factual: “Who wrote 1984?” → No external search needed.
   • Dynamic/Time-Sensitive: “Who is the current president?” → Requires live lookup.
   • Personalized Queries: “What’s my last order on Amazon?” → Needs user-specific data.
   • Opinion-Based: “What are the best sci-fi books?” → May require diverse sources.
2. Named Entity Recognition (NER)
   • AI detects if a query mentions specific brands, people, or events.
   • Example: “Tell me about OpenAI’s latest research.” If the query includes “latest,” AI is likely to use a grounded search.
3. Temporal Analysis
   • AI scans for time-sensitive language (e.g., “current,” “latest,” “as of today”).
   • Fast-changing data (e.g., stock prices) typically trigger grounded searches.
4. Confidence Scoring (How Sure Is AI?)
   • AI assigns a confidence score to its internal knowledge.
   • If confidence is high, it relies on pre-trained data.
   • If confidence is low, it triggers a web 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:

• Provide outdated or incorrect information about your business.
• Fail to recognize your brand entirely.
• Favour competitors whose data has been included.

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

1. Get Included – Ensure your brand’s content is part of AI training datasets using Waikay.
2. Maintain Accuracy – Keep your content structured and factually clear to prevent AI misinterpretations.
3. Monitor AI Outputs – Regularly check how AI platforms represent your brand and correct inaccuracies.

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