The good folks at Sonata Insights recently analyzed Datos‘ clickstream panel to uncover consumer search behavior insights in the “age of AI”. If you’re in the digital marketing world, it’s probably worth your time to download and read. Last Friday, I got access to the report and asked Datos for permission to share a few of the big findings and charts. Have a look:
![](https://images.sparktoro.com/blog/wp-content/uploads/2024/08/Visits-per-month-desktop-Google-vs-AI-1024x576.png)
This is a fairly solid, apples to apples comparison of traffic. It shows the visits/month on desktop to Google vs. the major players in AI/LLM search. In May of 2024, Google has slightly more desktop visits than they did in May of 2023 (+1.4% growth), while the largest of the AI search players by this metric (Perplexity), was at 15.69 (+42% growth from the year prior).
You might be thinking Google’s gone from 16X Perplexity’s size to only ~11X, but… this is only showing visits, not searches. Sonata and Datos graphed that, but removed the axis details, unfortunately:
![](https://images.sparktoro.com/blog/wp-content/uploads/2024/08/Monthly-active-search-users-by-platform-1024x576.png)
However, they did include this key data point: Google had more than 290x the number of search users as Perplexity in May. And compared to Google’s ~200 “events” (i.e. searches) per searcher per month, Perplexity is ~15. In essence, Perplexity’s users are about 7.5% as active on the platform as Google’s users are on Google.
That’s not to say AI search isn’t growing! It is. And more rapidly than the much more mature Google search engine. But, Perplexity is similar most similar in traffic, searches, and searches/searcher to Instacart or Kroger. As of right now, DuckDuckGo looks like a far more interesting potential competitor. I suspect the only reason they don’t get the same media attention is that we’re deep in the AI/LLM hype cycle, and headlines are far more likely to draw interest if they include some mention of an AI-centric product.
One of the ways I like to compare interest in my circles/network (where LLM/AI hype absolutely dominates my feed, especially on LinkedIn) is to compare it to mainstream, mature topics with steady interest on Google Trends.
![](https://images.sparktoro.com/blog/wp-content/uploads/2024/08/google-trends-ai-search-vs-seo-and-dinosaurs-1024x610.png)
Interest in AI search engines is absolutely big and growing. But do the topics get as many searches as a similarly-B2B subject like SEO? Or a mature, mainstream consumer topic like dinosaurs? Not yet.
Datos and Sonata’s report did have a few other findings worth mentioning:
- Google’s share of the combined desktop and mobile web traffic for the properties in our study remained essentially the same between May 2023 and May 2024. The percentage of users who searched was also relatively unchanged over the course of the year.
(to me, this means that even though some AI platforms are getting engagement, they’re not taking search share away from Google; if anything, it’s additive) - The average number of Google searches per desktop search user is rising, reaching 109.9 in May 2024, up 10.4% from May The number of searches per mobile web search user stood at 51 in May 2024.
(that doesn’t include the Google mobile search app, where a massive amount of mobile search takes place; Google themselves note mobile is ~2X the size of desktop) - The percentage of traditional search engine users who also use AI platforms has grown only 2 points in the past year. In May 2024, 16.45% of this cohort used at least one of the AI platforms in our study.
(that’s not a terrific growth rate for a supposedly massive, world-changing technology; we’ll see if the rollout of SearchGPT can change the trajectory, but I’m skeptical) - Importantly AI platform users didn’t give up on their old methods: 99% continued to use traditional search engines.
(Yikes. That does not match up at all to the dozens of posts in my social feeds claiming that Perplexity, Claude, ChatGPT, or Gemini has replaced Google entirely. Maybe I’m just following too many people in that weird 1%?)
I encourage you to check out the full report and show your clients, teams, and management next time they ask what your plan is for “marketing in the age of AI.” Apart from doing a healthy amount of PR and brand marketing (so your brand name frequently shows up in the sources LLMs are likely to use as training data), I’d argue 99% of your efforts should go to sources of influence that already impact your audience and customers vs. those you theorize might one day impact them. As we saw with the rise (and plateau) of mobile apps, the rise and fall of blockchain and NFTs, and the brief flirtations with the Metaverse, Augmented Reality (AR), and social audio (Clubhouse, et al), it’s far wise to wait until a platform is truly influential to your customers before trying to do marketing there.
As we always tell folks who use SparkToro—go where your customers are. Market to them in the places where they pay attention. If and when their behavior changes, follow them, but don’t try to predict where they’ll be in 2-5 years. In our opinion, there’s just not enough benefit to being influential in a place/on a platform before your customers get there.
p.s. If you’re looking for a place where your customers almost certainly are, and your competitors probably aren’t yet, it’s Threads. After passing 175M MAUs in July, they just crossed 200M in August, now on pace to surpass Twitter in <2 years. My account is regularly getting 10-50X the engagement I receive on Twitter ¯\_(ツ)_/¯