Rankings
Interactive rank tracking with unlimited competitors.
Content
AI insights and GenAI writer meets SEO performance.
Technical
Crawlability, indexation, rankings, traffic, etc. across URL(s).
Research
Instant topic, content, and competitor gaps.
Analytics
Where rankings, traffic, and business impact meet.
Track Visibility
Track brand citations and mentions in AI search.
Research Prompts
Uncover search demand for prompts, questions, and topics
Analyze Sentiment
Understand brand tone in AI results
Optimize Content
Create for AEO and GEO visibility.
Measure Performance
Track AI search traffic and conversions.
Discover Bot Activity
Know if AI bots access your pages
Monitor Accuracy
Detect hallucinations and errors in AI responses
Benchmark AI Visibility
Instant panoramic view of your AI search visibility
Accelerate Indexation
AI engine discoverability for every new page
Monitor Web Mentions
Amplify brand signals for AI search authority.
By Mitul Gandhi, Chief Architect & Co-Founder of seoClarity
Updated: January 9, 2026
Over the last year, we’ve been intensely focused inside ArcAI on one job: helping enterprise SEOs answer a deceptively simple question:
“Where is the real demand in AI Search?”
Today, we’re rolling out the latest version of our AI Search Demand Estimation model. It’s a more refined hybrid approach, built from hard-earned lessons, better calibration data, and a commitment to transparency that we think the entire industry needs right now.
But before I tell you what’s new, I want to walk you through how we got here, because the “how” is the part that builds trust.
Search demand has always been one of the most valuable signals in SEO. It’s how you prioritize:
That part hasn’t changed.
What has changed is the shape of demand.
In AI Search, queries aren’t short head terms. They’re long, multi-part prompts. They can include context, constraints, follow-up questions, and even preferences.
So if you’re trying to size opportunity, you immediately hit two constraints that everyone in this space runs into:
And if you’re an enterprise SEO, you can’t work with “hand-wavy.” You need something you can defend.
We started where our data advantage is real.
We’ve been collecting clickstream data for years, and it includes 1B+ new user prompts / questions / keywords each month. That’s an enormous sample of how people search and what they ask.
Our first model leaned heavily into that dataset:
This gave us something important early: directional signal. And it helped us build the foundation of ArcAI’s prompt research workflows.
In the first half of 2025, we learned fast that clickstream-only approaches break down when you try to serve enterprise needs at scale.
When we rolled this out across 3,500+ brands, we saw gaps that mattered:
If you’ve ever looked at a platform showing millions of searches for an obscure query, and then compared it to what Google reports, you know exactly what I mean.
This is the dirty secret of clickstream in general:
Clickstream can be huge and still be biased.
Bias in panel composition, device coverage, region coverage, vertical skew… all of it shows up eventually.
And if your job is to guide millions of dollars of content investment, “eventually” is too late.
So we did what we always do when the data doesn’t hold up:
We rebuilt the approach.
One insight changed everything for us:
Just because ChatGPT exists,
people’s problems haven’t changed.
The WHAT hasn’t changed. The HOW and WHERE have.
That premise became the anchor.
People still want the best running shoes, the cheapest flights, the right compliance framework, the top CRM-whatever your world is.
What’s changing is:
So instead of trying to “invent” demand, we asked:
What if we use the most reliable base we have for intent demand, traditional search, and model the shift?
Google still represents the majority of global search activity, and it has something AI engines don’t provide: a mature demand baseline by topic.
So we moved to a hybrid approach:
That’s the core model architecture that got us much closer to “enterprise-grade.”
Then something big happened.
A National Bureau of Economic Research paper (“How people use ChatGPT”) introduced a level of behavioral grounding that simply wasn’t available before. In plain terms: it gave the market a better window into how AI tools are being used, not just that they’re being used.
For us, that mattered because it let us do what every model needs at some point: stop guessing and start calibrating.
So we plugged that data into our hybrid framework and updated the adjustment layers accordingly.
And because this entire space is full of black-box metrics, we made a decision that’s a little unusual: We shared the methodology publicly.
Not because it’s “nice.” But because enterprise SEO needs defensible numbers, and the industry needs standards.
The version we’re rolling out now is the most refined iteration of that hybrid model.
Here’s what you should expect (and what we optimized for):
AI prompts don’t map cleanly to one keyword. This release improves how we distill prompts into:
(If you’ve used our prompt research capabilities, this is the same philosophy: don’t chase every variation, own the underlying intent.)
The clickstream-only model struggled here. The hybrid model improves consistency by grounding estimates in a stable baseline and applying calibrated adjustments, so you don’t get wildly inflated demand in one market and undercounting in another.
If you can’t explain it, you can’t defend it.
This release is designed to be “boardroom explainable”:
Because “AI Search Volume” is about to become one of the most abused metrics in the industry.
It’s easy to publish a number. It’s much harder to publish a number you’d be willing to defend in front of:
If AI Search is going to be a real channel (and it is), we need demand metrics that don’t feel like magic.
That’s also why Clarity ArcAI is built as an end-to-end system: prompts, visibility, optimization, performance, and measurement have to connect… otherwise “demand” becomes a vanity metric instead of a planning input.
This model is live because we believe it’s materially better, but we also know the fastest way to improve it is to put it in the hands of enterprise teams who live in the nuance.
If you’re already using ArcAI’s prompt research and visibility tracking, this update should make prioritization sharper and the “why this topic” conversation a lot easier.
Looking for an AI search solution to help prioritize AI efforts? Schedule a demo of seoClarity's Clarity ArcAI where you'll see first-hand how our end-to-end AI solution is helping brands turn AI search into a real channel for their organizations.
.png?width=192&height=192&name=Mitul%20Gandhi%20-%20Author%20Snippet%20(1).png)
About the Author: Mitul GandhiAs a longtime data-driven serial entrepreneur, information architect and SEO veteran, Mitul has developed a blend of vast technical expertise and intense marketing insight. His variety of experience, gained in positions in in-house SEO, search marketing, and software development, affords him the ability to efficiently assess how to use software tools to meet challenges and drive ROI. As the Co-Founder and Chief Architect of seoClarity, Mitul currently oversees day-to-day operations, and provides strategic direction to all departments. His well of knowledge includes 10+ years of consulting experience with Fortune 500 and top Internet retailers concerning online search marketing. He has several patents pending for analyzing cause and effect in SEO. Mitul holds an MBA in direct marketing from Rochester Institute of Technology. Additionally, he has spoken at conferences in the United States and the U.K., including SES, SMX and Pubcon. He has also been quoted in MSN Money, USA Today, Time Online, Search Engine Watch, Search Engine Land and Web Pro News. Connect with him on Twitter or LinkedIn.
