Insights • 7 min read

35% of our demo sessions looked broken... from email scanners

Nick Laudato

Sr. Manager, Manager, Web Strategy & UX Development at Fullstory

Last updated: 11/20/2026

Table of Contents
  • The two prompts
  • Enter link scanners
  • What it changed
  • Get more from StoryAI
  • Use AI as a thought partner
  • Featured
  • Return to top

Article summary

I used StoryAI's Issue Investigator to review technical issues on our demo form page. The biggest was a React error in 1,607 sessions, about 35% of demo page sessions. A follow-up prompt showed the traffic was bot clicks from email security scanners, not prospects, so the fix is in my segments, not the page.

Key takeaways

  • The top "issue" on our demo page was bot clicks from email security scanners, not prospects.

  • One follow-up prompt turned an alarming error count into a confident "don't fix this."

  • Treat AI as a thought partner: scope the prompt, then pressure-test the answer.


For a second, I thought something on our demo page was broken.

I had asked StoryAI to look for technical issues that might stop people from filling out forms, and the top result was a React error on /demo, hitting 1,607 sessions. Of all the traffic that landed on our demo page in the last 30 days, 35% hit this error. That page, /demo, is one of the most important pages on our site, so my first reaction was simple: something is wrong.

Well, not for real visitors, it turns out. I had a different answer one prompt later.

Why I run a CRO sweep every week

My work on fullstory.com spans SEO/GEO and website strategy, and our demo form sits at the center of it. It's our main source of lead generation. If something goes wrong there, it's the most impactful thing to fix. It ties directly to our AQL goals (our version of marketing-qualified leads, or MQLs) and pipeline.

I try to do a conversion rate optimization (CRO) sweep every week to get a feel for our funnels. I'm a firm believer in regular technical checks, and with Q4 here, they matter even more. So I often start by asking StoryAI.

Before StoryAI, I would have analyzed everything manually. It’s too slow, and there's so much data. Don’t get me wrong, I still love digging into sessions. Session replay helps me experience what the user experiences, even if it's a small percentage of them. But the data is already in Fullstory, so why not offload that analysis?

Prompt 1: scope it tight

My first prompt to StoryAI looked like this:

"On hostname = www.fullstory.com, for the last 30 days, identify technical issues that may cause users not to fill out forms on pages including /ps/ or /demo. I'm looking for CRO opportunities."

storyai-agents-1 1

Our paid search pages include /ps/ in their URLs, which helps with attribution and analysis. Our main demo page is /demo.

I wrote the prompt on the fly, but I know what helps StoryAI hone in.

  • The hostname needed to be clear that it's our marketing site.

  • I wanted a large enough time range to gather more insights.

  • I wanted a clear page or two that had our form or form events on them.

StoryAI scoped the analysis to our marketing segment that excludes internal and test traffic, then returned the results.

At the top was a console error: "page resources for /demo/ not found. Not rendering React." It hit 1,607 sessions.

Prompt 2: lead with a hypothesis, then test it

I was shocked. The number was jarring, and I had never seen that many sessions hit a single issue. But our forms were still converting as expected, so something didn't add up. I went in with a hypothesis that this might be bot traffic.

"Dig into the first issue, the page resources and React issue. Identify trends of users: device type, geographic, IP, city, etc. I want to understand whether this is bot traffic or real issues with users on our site."

storyai-agents-2 1

StoryAI routed this to Issue Investigator, one of its Expert Agents built to dig into a single problem and find the cause. I asked for specific attributes that would help narrow it down. The prompt wasn't polished, but this was a test of my hypothesis, not a fishing trip.

Enter link scanners

Of the 1,607 sessions, 1,606 were desktop. Every sampled session ran Windows, and the browsers showed headless or automation signatures. The traffic came from concentrated data center hubs and cloud gateway nodes, with Dublin, Ireland (AWS and Azure) at the top of the list, followed by US hubs like San Jose, Quincy, Boydton, and Dulles.

The source of traffic was the first thing that jumped out to me. We see traffic from lots of specific hub regions, to the point where I've filtered some of them out of certain views. Put Dublin together with all Windows and headless browsers, and it was clear there was no way this was just real people hitting this issue.

User behavior told the same story. StoryAI found zero sessions across the entire cohort where a human actually interacted with form elements. The bot clicks followed a predictable pattern: open the page, wait 25 to 65 seconds, disconnect.

StoryAI's Issue Investigator concluded that these were email security scanners, the tools that open email links before your marketing emails reach an inbox, to make sure none is a phishing link. Nearly all the volume reflects automated security link pre-scanners running in sandboxed Chrome instances.

storyai-agents-3 1

I knew scanners like this existed, but because they look like ordinary Windows desktop browsers, my normal bot exclusions didn't quite catch them out of the box.

Our email automations add a q_mailing_ parameter to email links, which flags any session that started from an email. I later asked StoryAI how many of the 1,607 sessions came in through a mailing link. It found 1,089 with q_mailing_ in the URL, which was about two-thirds.

StoryAI was right to flag the error. It was an issue. But was it one that affected pipeline or AQLs? I can say confidently that it was not.

What it changed

No wasted dev time. The error count said broken. A manual check would have said fine. Without StoryAI's context, we'd have burned dev cycles chasing a bug real visitors never hit.

A more accurate conversion rate. If we counted those bot clicks in our total conversion rate, we'd be understating it. Scanners aren't there to convert in the first place.

Clear next steps. StoryAI recommended excluding traffic tagged with q_mailing_* parameters from marketing segments. I'm moving this traffic into our bot and automated traffic segment so we can analyze it separately, and keep it out of weekly reporting views and future form analysis. Anything else that gets through will be added to that exclusion. StoryAI also recommended re-baselining form drop-off analysis around real intent signals, like mouse movement, scroll depth over 10 percent, or form field focus.

storyai-agents-5 1

How I get the most out of StoryAI (or any AI for that matter)

  • Scope the prompt. Do the setup work for the AI. Tell it the hostname, the time range, the event. The more you give it, the better. If you're missing a detail, StoryAI can usually fill it in.

  • Use your segments. I keep predefined segments that filter out internal users and test traffic. Next time I may go in with a specific segment in mind, but running broader is what surfaced this.

  • Work with it like a coworker. It's not going to do it all for you. Workshop it. Doing this has helped me understand our website as a product even more. Call it human in the loop: you're still the one orchestrating.

Use AI as a thought partner

The numbers said something was broken, but context said who was actually seeing it.

Pressure-test the answer. StoryAI wasn't wrong. It identified an issue, but I questioned the result and dug deeper. That is how I learned what not to fix.

And none of this works without the data underneath it. We wouldn't have had any of these insights without Fullcapture.

What's next for me

StoryAI's Expert Agents, Conversion Optimizer and Issue Investigator, are the bread and butter of my work right now. Our website is a sales engine, so any CRO opportunity helps improve the experience for our users. I'm taking the same approach to agentic traffic, figuring out which AI agents hit our site and what they do there.

If you want to try this out, StoryAI Agents are available with StoryAI Premium. Open StoryAI, start with the two prompts above, swap in your own pages and questions, and then dig deeper into a result. The follow-up is where I found the real story.

Want to see it in action first? Here’s a short StoryAI Agents demo.

Explore StoryAI Agents →

Google logoWant more like this in your Google results? Add Fullstory as a preferred source.
Sr. Manager, Manager, Web Strategy & UX Development, Fullstory

Nick Laudato leads web experience strategy at Fullstory, where he owns SEO, GEO/AEO, CRO, and site architecture. He has almost a decade of search experience across B2B SaaS, ecommerce, and agency work. He lives in Northeast Ohio with his wife and daughter. Outside of work you’ll find him fishing for smallmouth bass or burning through AI credits building things nobody asked for.

Additional Resources

Cartoon character Meta Muse holding a magnifying glass over a robot icon, with a analytics chart and Meta logo on a green background. Concept image for tracking AI Agent traffic.
How I track AI agent traffic on our website

How I separate AI referrals, crawlers and agents, then use Fullstory and MCP to see what agents do on our site.

Read the blog
fs-bash-blog-post-event
The trust gap in agentic analytics (and how we're closing it)

StoryAI Agents investigate on demand and around the clock, and every conclusion links back to the session replay behind it.

Read the blog
Illustration of a globe with a rising bar chart and an AI chip icon, representing global AI adoption statistics
58 AI adoption statistics for enterprise leaders

Most organizations already use AI, but far fewer can show that it's paying off. These AI adoption statistics break down who's getting results and why.

Read the blog
A smartphone with a loading icon, a stopwatch, a server stack, and a cloud, representing digital experience monitoring.
What is digital experience monitoring? From performance to user behavior

Digital experience monitoring (DEM) measures how your site or app performs for users. Here’s how it works and what behavioral data adds.

Read the blog
MCP-Ask-Watch-Loop
Ask. Watch. Loop. Get more out of Fullstory MCP.

Go beyond basic questions with Fullstory MCP. See how top users run agentic session review and scheduled loops to catch silent failures and size bugs.

Read the blog
fs-bash-blog-chris
What the Houston Oilers taught me about AI

Personal AI agents are changing how people shop and how teams work. Learn why judgment matters more than speed, and what agents need to get it right.

Read the blog