How ChatGPT's SearchGPT Will Shape Your AI Optimization Strategy

Google’s reign isn’t what it used to be. 

Gartner has forecasted a 25% drop in traditional search engine volume by 2026, with an expectation that Google will cede market share to platforms like ChatGPT.

While Google has maintained decades of dominance in search, we’re witnessing a slow decay in progress. For the first time since 2015, Google’s search market share dropped below 90% in Q4 of 2024.

Much remains to be seen about how users will engage with the internet for information discovery, but one thing is for sure: Marketers can no longer presume their bases are covered by optimizing for Google search.

2025 will be a year of learning how, when, and why to optimize new platforms.

Let’s start with ChatGPT.

The FLIP Framework

We believe that Freshness, Local intent, In-depth context, and Personalization are key factors in determining whether or not GPT will search the web to answer a question.

How does ChatGPT Search work?

SearchGPT was a prototype of ChatGPT’s search feature, which has been available to all logged-in ChatGPT users since December, 2024. Queries that require more up-to-date information than ChatGPT’s training data are supplemented with citations found on other web sources.

ChatGPT won’t always leverage SearchGPT. The system will choose if it should search the web based on what you ask, or you can manually force it to search the web by clicking the web search icon.

If a user simply asks a question, ChatGPT will decide if it should respond using its training data or will perform a web search and use that information to formulate a response and cite the sources used.

How does ChatGPT decide when to search the web and turn into SearchGPT?

While OpenAI’s GPT-3 was trained on 570 GB of CommonCrawl data, GPT-4o’s training went further, including text, vision, and audio as well as external domain experts, research institutions, and civil society organizations.

But, even with these improvements, the knowledge cutoff for GPT-4o models is October, 2023 (as of the date this blog was published) and, while the training data is comprehensive, it isn’t omniscient.

When a question requires information outside of the scope of its training data, ChatGPT will search the web.

An example of these pathways in action can be seen if you ask a question like “what are the best data visualization tools?” vs “what are the best data visualization tools in 2025?”.

The modifier “in 2025” is an easy way to force ChatGPT to search the web, since its training data only includes information through 2023.

However, we should always remember what we know about how people actually search. We know from Google’s Monthly Search Volume data that few users will include recency modifiers like “in 2025”. 

If we think about this in terms of keyword demand, "best social media tools" has 100 – 1K MSV while "best social media tools 2025" only has 10-100 MSV (Google Keyword Planner). In other words, most users won’t modify searches in a way that forces web search in ChatGPT.

Rather, we can presume that ChatGPT will still answer many queries with pre-trained dat vs using a web search and SearchGPT.

We tested this by taking some of the same questions from our previous study, like:

• What is the best online banking platform for e-commerce LLCs?

• Which bank has the best CD rates?

• What are the best tools to automate social media posts?

We asked these questions in the ChatGPT web interface, testing in both the free and paid versions, without tweaking any settings to emulate a typical user, and found:

In the paid version:
35% of the questions triggered a web search
65% of the answers came directly from training data

In the free version, it was similar:
30% of questions triggered a web search
70% relied on training data

What does this mean for you?

ChatGPT will likely rely on its training data rather than searching the web to answer some questions.

This means brands and marketers need two strategies:

1. Optimizing for questions that are answered using pre-trained knowledge
2. Optimizing for questions that are answered using web searches, such as those with recency modifiers (e.g. “in 2025”)

So how do you know which is which?

If you’re monitoring a smaller question set, you can always ask those questions directly to ChatGPT and see for yourself.

For larger question sets, we’ve created a framework to outline factors that generally trigger a web search.

FLIP Framework to winning in ChatGPT & SearchGPT

How to apply FLIP to priority search terms or questions you’re tracking

So, how can you use this framework to adjust your strategy? Let’s look at some examples from the perspective of the higher education industry.

Freshness

Question: “What are the enrollment trends for the upcoming fall semester this week?”
Sources GPT Might Use: University websites, government education portals (e.g., IPEDS, NCES), and campus apps with real-time alerts.
What you can do: Regularly update your institution’s website, internal data dashboards, and announcements.

Local Intent

Question: “Which campus library is open late tonight?”
Sources GPT Might Use: Campus maps, building hours, local event listings, and student life websites.
What you can do: Keep listings current for library hours, dining hall schedules, and campus bus routes.

In-depth Context

Question: “How many more credits do I need to graduate with a dual degree in Biology and Environmental Science?”
Sources GPT Might Use: Degree audit systems, research databases, and faculty portals.
What you can do: Centralize academic info, including curricula, prerequisites, and credit audits, in well-maintained online repositories.

Personalization

Question: “Based on my GPA and career interests, which electives should I take next semester?”
Sources GPT Might Use: Student records, career services platforms, and advising notes.
What you can do: Maintain accurate student profiles, including transcripts, progress trackers, and interest surveys.

We’ve introduced the FLIP framework, and now you’ve seen how it could work in the higher education industry. Whether it’s keeping your content fresh, addressing local intent, providing in-depth context, or leveraging personalization, the same principles can be applied to your industry. 

The way users engage with search will continue to change over time, while some may see this as a challenge, we like to look at this as an opportunity. The key is to adapt early, test often, and embrace the strategies that will keep your brand visible in search.

Want to stay informed on how to adapt your strategy to these shifts? Join our newsletter below to get insights delivered straight to your inbox.