Who Rules AI Search? What 3 Months of AI Visibility Index Data Tells Us

After recently launching AI Visibility Index, the definitive AI search study, we now have data for August, September, and October. 

This means we’ve been able to find deeper trends and track disruptions across ChatGPT and Google AI Mode, highlighting changes in source diversity, brand visibility, and vertical-level volatility.

AI search is rapidly evolving, but these observations, alongside the AI Visibility Index’s downloadable blueprint, put you in pole position for creating a winning strategy.

Some key questions we’ll answer in this analysis:

• Is Reddit still dominant as a source?
• Was there a ChatGPT algorithm change in September?
• Did Google’s removal of 100 results (num=100) per page affect AI search results?
• How stable are AI search rankings over time?
• How can brands react to these changes?

AI Mentions are Volatile and Differ by Model

Let’s look at the overall trends with mentions first. 

ChatGPT saw a roughly 12% increase in the number of brands mentioned across in September, before returning to a similar level in October. This suggests experimentation and possibly attempts to diversify prompt responses.

By contrast, the number of brands mentioned dropped by 4% in Google AI Mode from August to October, hinting that AI Mode has more defined controls on the recommendations it makes. 

For sources, ChatGPT saw an increase of around 80% in October. This indicates a fundamental shift towards using more sources, and a greater diversity of sources, in each response. 

Google AI Mode’s sources increased by 13% between August to October. This points to a similar, albeit lesser, shift.

Models Shows a Clear Split in Brand Diversity

ChatGPT offers a mixed picture, again demonstrating AI search’s unpredictable corrections. Most noteworthy:

• Consumer Electronics had a large jump, with over 20% more unique brands mentioned
• Technology & Software also saw a 12% lift
• YMYL category Finance witnessed a 15% drop

For Google AI Mode, it’s almost universal decline. Every vertical except Fashion & Apparel saw significant decreases in brand diversity (-10% or more). 

Key takeaway: ChatGPT and Google AI Mode have different approaches to which brands they mention. Each model needs to be tracked individually.

Note: We calculate diversity as the number of unique brands mentioned per total prompts in each industry.

Changes in Source Diversity are Equally Unpredictable

Volatility is the word for source domains as well. 

In ChatGPT, source diversity increased drastically and uniformly across all categories. The data shows similar shifts across the board, almost doubling the number of sources in most verticals.

Meanwhile, overall source diversity remains fairly stable in Google AI Mode. 

Key takeaway: At the moment, sources are stable in Google AI Mode, but ChatGPT showed huge changes in October. This volatility illustrates just how fast moving AI search can be.

How Stable are the Top Performing Brands?

Over three months, there were 25 new entrants in the top 100 brands ranked by weighted Share of Voice. The majority were in later positions, with just two of these were in the top 50. 

This tells us that the top 50 are fairly stable over time, but the volatility in AI search is enough that the bottom half of the top 100 can change.

61 of the top 100 brands were common in both ChatGPT and Google AI Mode, showing a high degree of similarity.

Note: To assess the performance of brands across both models we use a metric called weighted Share of Voice. 

Share of Voice takes into account the position a brand is mentioned, how often it’s mentioned, and how popular the prompts it appears for are. Based on user numbers, we give 80% weight to ChatGPT and 20% to Google AI Mode to come up with a combined score.

What Can We Learn from the Top Performers?

It’s stable right at the top. Looking at the top 15 performing brands overall, we see that the change in SoV is within the +/- 20% range. 

This means that top brands shouldn’t expect to gain or lose greater than 20% in AI visibility over a similar period. Comparatively narrow to the wider shifts.

Note: These changes are measured as a percentage change. This is a fairer comparison that allows us to compare brands of different sizes.

Source Domain Changes in Both Models

There are significant differences in both the number of sources and their frequency between ChatGPT and Google AI Mode, therefore we’ll look at each AI model individually. 

What Are the Top Sources in ChatGPT?

It’s that word again: volatility. The average change that domains saw in prompt coverage is around 120% across the top 100. 

For the top 10, the largest change is Reddit dropping by about 82%, which is a significant correction. 

Note: Here we measure in prompt coverage %. This is how many of the prompts a particular source domain features in. For example, if ChatGPT uses your domain as a source in half of the prompts, your result would be 50%. 

Overall, we see Reddit used as a source in 80% fewer prompts. That being said, our first data from August found Reddit being used as a source at an exceptionally high rate: 135% of prompts. This means it was used an average of 1.35 times for every response. 

Therefore, the change seems to be a natural correction. 

It means Reddit is still used in around 24% of prompts, and sees the domain positioned as the fourth most used source in ChatGPT overall. This is behind Wikipedia, Forbes and Amazon.

What Are the Top Sources in Google AI Mode? 

Google AI Mode sees highly significant changes in October compared to August. 

Contrary to ChatGPT, the most noteworthy change here is Reddit’s increased usage as a source—which is a fascinating development. 

Reddit was cited 75% more and was used in 20% of answers. This makes Reddit the second most used source in Google AI Mode, behind only Amazon.

Other significant shifts include increases for Yahoo and Best Buy. 

Yahoo’s gain chiefly comes from Yahoo finance. Clearly this domain has become a trusted source for Finance in AI Mode. 

Best Buy is an interesting case. These gains are split between user reviews and product pages, serving two AI search needs in UGC reviews and factual product info.

Note: Google regularly tests linking to its own properties, like Google Shopping, within AI Mode. These tests make fluctuations in its source frequency natural and we can expect them to continue until Google has a “standard” version of AI Mode. 

Are the Brands that ChatGPT and Google AI Mode Mention Becoming More Similar?

The overlap has remained consistent and fairly high, with very little change between ChatGPT and Google AI Mode across verticals. Industries do see different extents of overlap though, with Consumer Electronics sharing the most brands between the two models and Business & Professional Services being the least similar.

Key takeaway: Brand similarity is fairly high and remains stable. While the approach to become a top brand may vary by AI model, once you’re there, you’re hard to shift.

Are Sources that ChatGPT and Google AI Mode Use Becoming More Similar?

Overall no, there’s less overlap in October than in August. But this also depends on the vertical. 

Sources used in the Fashion & Apparel and Business & Professional Services verticals have become more alike, but the changes are quite low: 

Key takeaway: Source similarity is much less than brand overlap. Therefore, you’ll need to take a varied approach for each AI model and tailor this to your industry. 

Reddit Was Corrected in ChatGPT but Arrived in AI Mode

Perhaps the biggest change was Reddit’s recalibration as a source in ChatGPT. This chart shows the complete picture across verticals and months. 

It would be fair to say that Reddit was overrepresented in August and ChatGPT has scaled this back to more expected levels in September and October. 

Overall, Reddit remains vital. It’s still used as a source for 15-33% of prompts depending on the vertical (and remains a top 10 source in each).

During the same time period, the opposite is true for Reddit in Google AI Mode.

Reddit saw uniform gains between August to October, with its use as a source overall increasing from 11% to 20% of prompts. 

This is close to ChatGPT’s overall usage (24% of prompts), meaning that the two models are beginning to converge on Reddit usage. Let’s see if this trend holds true. 

Explore all of these findings for yourself in the AI Visibility Index’s interactive leaderboards and visuals.

TL;DR: The Key Questions, Answered

Is Reddit still a key source in AI search after ChatGPT’s correction?
Absolutely. Reddit remains a top 10 source in all verticals and the fourth most cited source in our dataset. Alongside this, Reddit is gaining prominence in Google AI Mode, now becoming the second most cited source.

Was there a ChatGPT search algorithm change in September/October?
It seems likely ChatGPT tweaked how answers are composed, seeking to use more sources per answer.

The extent of Reddit’s correction seems manual. However, other changes, like the overall number of sources increasing hugely, are significant enough to support this theory. 

Did Google’s removal of seeing 100 results per page affect AI search results, and if so how?
Both the number of sources and mentioned brands in ChatGPT increased since this change. If its crawling ability was suddenly limited to only domains ranking in the top 10, we would expect more significant shifts in the data. 

While the Reddit scaleback is significant, this seems more likely to be an internal change, than directly related to Google.

How stable are AI search rankings over time?
Fairly stable. We saw 25 new brands enter the top 100, but mainly in the lower positions. A high degree of overlap is also observable between the models. 

How should brands react to these findings? 
ChatGPT and Google AI Mode agree 67% of the time on which brands to mention, but just 30% of the time on what sources to use. 

This means that you need to go after media mentions using an AI model specific strategy: the sources ChatGPT uses are fundamentally different from those that AI Mode uses. To maximize your audience reach, you need to do well in both. 

Semrush Enterprise AI Optimization lets brands track thousands of prompts across all leading LLMs, then identify which sources are most important and how to optimize for them.

Explore the AI Visibility Index for More Insights

Explore our AI Visibility Index interactive site to discover even more trends, decode the tactics that work, and stay ahead of the competition.