Jerry.ai
US car-insurance super app
How Jerry more than doubled its conversions from AI search in eleven weeks
Using Sitefire, Jerry more than doubled the customers that AI search sends it, in eleven weeks. Jerry used Sitefire’s data to improve the web pages it already had, so that AI answers would cite them and send buyers to jerry.ai.
Three of these four numbers are real behavior: AI models pulling pages to build a live answer, people arriving, people signing up. The fourth is a visibility score, measured against a tracked sample of buyer questions. This is the first case study in AI search to report all four.
2x+
monthly conversions from AI search
People who found Jerry through an AI answer and became customers, over the eleven weeks after the improvements went live.
+37%
AI referral traffic
Real visits from people who clicked through from an AI answer.
+60%
AI bot traffic
How often AI models pulled Jerry's pages to build an answer to a live question.
+29%
AI visibility
Jerry's standing across a tracked sample of the questions its buyers ask.
The work
Jerry used Sitefire to find its biggest opportunities, and shipped them
Jerry ran one loop with Sitefire, over and over: find the questions its buyers actually ask, see which of its pages the models read and where competitors get cited instead, then improve those pages and publish.
Jerry shipped several waves this way across the spring, each one identified, drafted and live with little lift from the team.
For each page, Sitefire showed what the top-cited content in that category had in common, and what the models reward. Jerry’s writers built that into the pages systematically.
How it worked
1
Diagnose
Sitefire’s agents map which questions buyers ask, which pages the models read, and where competitors get cited.
2
Optimize
Using Sitefire, Jerry improved its pages for AI search - one prioritized brief per page, naming exactly what to add.
3
Measure
Standing on the tracked questions, and what real traffic and signups did afterward.
“Sitefire showed us which pages actually mattered for AI Search and what each one was missing. The goal was never to publish more - we already have a pretty strong content production engine. Instead, we wanted to make the pages we already had get trusted and cited by the LLMs.”
Ida Sultan
Chief of Staff, Jerry
The results
All four signals up after the improvements shipped
The four charts track each signal week by week, across the engagement, each shown against its own level before the work went live. Visibility Score is the share of tracked buyer questions whose AI answers mention jerry.ai. Citation Share is jerry.ai’s share of all the citations in those answers. Both are defined at sitefire.ai/docs/kpis.
Jerry optimized for AI visibility. Traffic followed.
Sitefire scores AI visibility against a tracked sample of the questions Jerry’s buyers ask. A sample stands in for a population nobody can observe directly, so it is worth exactly as much as its likeness to the real thing.
Jerry improved the pages those tracked questions pointed to, and the score rose. Then the real population moved the same way: AI models pulled those pages more often to answer live questions, more readers arrived, and more of them became customers.
Real behavior is the only thing that can confirm a sample was a good likeness - which is why Sitefire reports it next to the score.
Conversions from AI search2x+
per week, against the week before the improvements went live
AI bot traffic+60%
per week, against the week before the improvements went live
Visibility Score+29%
share of tracked buyer questions mentioning jerry.ai
Citation Share+12%
jerry.ai's share of all citations in those answers
The within-site control
Only the pages Jerry improved returned above where they started
To isolate the effect of the work itself, we compared the pages Jerry improved with Sitefire against the rest of jerry.ai over the same weeks. Both groups sit on the same site and face the same competitive conditions, so the only real difference is the improvements.
Weekly AI bot traffic, improved pages against the rest of the site
each group indexed to its own level before the March shock
Each line shows a group’s weekly AI bot traffic against its own level before March. Both dropped when GPT-5.4 launched, and both recovered, but not equally. The improved pages averaged 112% of their old level and ended above where they started, while the untouched pages averaged just 72%. The median untouched page lost 38% of its AI bot traffic over the period, and twice as many improved pages grew as untouched ones.
That gap is the case. The market moved against everyone. Only the pages Jerry worked on came back.
In short
In short
Using Sitefire, Jerry turned the content pages it already had into the ones AI models rely on. AI bot traffic grew 60%, AI referral traffic grew 37%, and monthly conversions from AI search more than doubled.
The visibility score rose too, and the order matters. A score moves whenever you optimize against the sample behind it. What it cannot show on its own is whether that sample looks like the people you are trying to reach. Here the real population answered, and it moved the same way.
Sitefire also tracks which of your pages AI models read, and which page carries the link a reader can click. They are often not the same page, which is why standard analytics credits the wrong content.
What comes next
Jerry and Sitefire are now building the measurement that is still missing, to drive the next round of growth. It connects the pages AI models read to the signups they produce, so the next improvements go to the pages that actually grow the business.
“Sitefire is the first tool we have used where the analytics go past synthetic tracking, and recommendations actually work.”
Ida Sultan
Chief of Staff, Jerry
Methodology
- Windows. Headline changes compare the last full week before the improvements went live with the most recent full week, the same window for all four metrics. Weekly charts run from onboarding in February through the end of June.
- Relative reporting. Every figure in this study is a change or a share. No absolute volumes appear anywhere, and no chart axis carries one.
- Two kinds of measurement. AI bot traffic, referral traffic and conversions record real events: models fetching a page, a person arriving, a person signing up. Visibility Score and Citation Share are scored against a fixed sample of buyer questions and describe standing within that sample. The two are reported separately throughout and never combined.
- AI bot traffic counts the models that read a page to answer a live question. The controlled comparison sets the pages Jerry improved against comparable untouched pages on the same site, over the same weeks.
- AI referral traffic is a floor, not a ceiling. Many AI-influenced buyers research inside the model and then navigate directly, so they never appear as referrals at all.
- The tracked question set was fixed at onboarding and never changed afterward. Holding it steady is what makes the later comparison with real traffic meaningful (definitions at sitefire.ai/docs/kpis).
Get started with Sitefire
Take action to win customers in AI Search
