By There is a lot of activity moving to LLMs and generative AI tools right now, and for marketers, founders, and creators, the questions are almost always the same:
- Are my customers moving from traditional methods of discovery to these platforms?
- If so, how much?
Starting today, SparkToro has the answers.
Using clickstream panels from our partners at Datos, we’re able to see when, where, and how much people who visit a particular site, or have specific search behaviors/profile characteristics use these discovery platforms. And we’ve broken that information into a specific report called “Search & AI Tools.”
Like our very popular “Social Networks” tab, the “Search & AI Tools” has three columns of data:
- Audience Affinity – on a scale of 0-100, how relatively popular is this discovery platform with the audience you’ve searched for
- US/UK/CA Affinity – depending on the country you’ve chosen to search, this column will tell you how globally popular a discovery platform is with all audiences in that geography
- Affinity Change – my favorite column, this one shows the delta between your audience and the country’s overall population. Sorting by it can give you a quick view of the discovery tools your audience is using more/less than most.
Now, a few VERY IMPORTANT caveats on this data:
Caveat #1: Affinity is not a linear scale. Don’t misinterpret Reddit.com, a very popular discovery platform at 75/100 affinity with US web users, to have 75% of the visitors of Google.com at 99/100. It’s far less than that (SEMRush estimates Google’s traffic at ~132B visits/month and Reddit at 5.7B, though I suspect both of these are high).
Caveat #2: Affinity Change, while an excellent way to see what’s more/less popular with your audience than average is relative, not absolute! For example, if you look at the discovery platforms that are popular with paralegals you’ll notice that Consensus (5/100) and Startpage (3/100) have very high Affinity Change scores, which is interesting and potentially useful, but… relatively speaking, they’re still used a fraction as often as ChatGPT (at 32/100).
Sorting by Affinity Change is a powerful way to quickly spot what’s uniquely popular with your audience, but make sure you compare it to the overall Audience Affinity score so as not to be biased by big changes among small numbers.
Caveat #3: Affinity scores are relative to their respective sections in SparkToro. If you see a website in the “popular sites” tab with a 75 affinity score, that’s relative to the other sites in that section, not to the affinity numbers in the Search & AI Tools tab (if we made it the same throughout SparkToro, Google, YouTube, Faceboook, and Instagram would be in the 80-100 and most everything else would be 1-5 😅; people visit those big sites A LOT).
Caveat #4: OpenAI and ChatGPT both appear in this list and will for the next few months at least. That’s because our clickstream data is inclusive of traffic patterns from the transitionary period when OpenAI launched and moved most ChatGPT traffic from OpenAI.com to ChatGPT.com. Over the next two quarters, expect to see ChatGPT.com become much more prominent and OpenAI.com shrink. For now, I’d recommend treating them as a single entity (unfortunately, we can’t combine the two domains in our data models without big, frustrating problems).
How to use this data:
My recommendation is to run an analysis for your own site (or your client’s, if you’re doing consulting work) AND several competitors, first. You can then build custom charts like I did below for various popular CRM tools:
It shouldn’t be surprising that Search & AI tool affinity scores for these sites are similar given that they attract quite similar audiences, but it was still a nice validation of our data!
A graph like this might help you make the case for a marketing investment in LLM visibility to leadership (at least, at a CRM company). Side note: we’ve got some advice on how to get your brand into LLMs and AI tools here.
Then, I’d look at individual groups you’re trying to serve (job titles is a good one, search keywords is another). For example, I analyzed the audience that has “holistic medicine” in their profile, and found a surprising affinity for traditional, old-school search engines (Bing and Yahoo), as well as privacy-centric search tools like Brave and DuckDuckGo.
Since LLMs are relatively new, growing rapidly, and have a lot of non-discovery use cases (our estimate is that only ~10% of ChatGPT prompts are replacements for traditional types of search activity), visitation affinity isn’t always a sign that your audience is switching away from Google. We don’t know exactly what any given user does once they’re on these website, but we’ve got the data to prove when they’re getting more/less visitation than usual from a particular audience.
As always, we’d love your help putting this new feature through the ringer; drop us a line to [email protected] with any feedback or suggestions.