Client result · anonymised · real GA4 + Mixpanel data

GEO case study: how a SaaS app grew AI-referred signups 8X

A product-led SaaS app was almost invisible in ChatGPT and Gemini, and its analytics could not even see the few visitors AI sent. We fixed the measurement, the crawlability and the content. In the first 16 days of November it got more AI-referred signups than in the previous six months combined.

8XAI-referred signups1-16 Nov vs an average month, Jul-Oct (Mixpanel)
4.6XAI assistant sessions a day1-15 Nov vs 15-30 Sep (GA4)
2.8Xsignup rate vs Google organicAI search visitors, last 6 months (Mixpanel)
2.6XGoogle AI feature impressions90 days vs the previous 90 (Search Console)
The starting point

AI assistants were sending buyers. Nobody could see them.

The client is a self-serve SaaS product with a free signup and paid plans. Search was its second channel after paid ads, and its team wanted to know one thing: are ChatGPT and Gemini recommending us, and does it turn into signups?

  • Blind analytics. The tracked channel field ignored UTM tags, so paid ad clicks were reported as organic and AI assistant visits sat in Direct.
  • Pages AI could not read. Key pages sent their canonical tag to an app screen, and many pages served bots an almost empty shell.
  • Content built for keywords, not questions. Single posts tried to rank for pricing, free plans and comparisons at once, so no page gave AI a clean answer to quote.

Engagement scope: technical SEO, content, internal linking, analytics and GEO. Programme start: September 2026.

Result 1
8X
AI-referred signups, 1-16 November vs an average month (Jul-Oct)

AI-referred signups were roughly flat for four months, then jumped

From July to October the client averaged the same small number of AI-referred signups each month. In the first 16 days of November it recorded 8X that monthly average, and 1.7X as many as May to October combined.

The jump came after the canonical fixes, the new comparison and answer pages, and the internal linking work were all live and recrawled.

AI-referred signups by monthIndexed: average month Jul-Oct 2026 = 100. A product-led SaaS app, anonymised.
49May
31Jun
92Jul
80Aug
111Sep
117Oct
8001-16 Nov
Source: client Mixpanel, Top Search Agency channel model (AI Search = visits from ChatGPT, Gemini, Claude, Perplexity, Copilot and others)
Result 2
4.6X
AI assistant sessions per day, 1-15 November vs 15-30 September

More AI visits, and more of them doing something

Sessions from AI assistants rose 4.6X per day in two months. Key events from those sessions, such as starting a signup, rose 6.8X, faster than the visits themselves.

AI assistants went from 1.8% to 4% of all sessions. Still a small channel, but its share more than doubled in two months.

AI assistant sessions and key events, per dayIndexed: 15-30 September = 100. Per-day rates, so part months compare fairly.
100100
15-30 Sep
191206
October
463676
1-15 Nov
AI assistant sessions per dayKey events from AI sessions per day
Source: client GA4, AI Assistants channel group built by Top Search Agency

Is AI search sending you buyers, or hiding them in Direct?

Our free audit checks AI crawler access, the buyer prompts that name you or your rivals, and how your analytics count AI traffic.

Result 3
2.8X
signup rate of AI search visitors vs Google organic, last 6 months

AI visitors arrive ready to sign up

In Mixpanel, 35.6% of AI search visitors signed up, against 12.7% from Google organic. GA4 tells the same story with its own funnel: 53% of AI assistant visitors reached the sign-up step, against 11% from organic.

They also engage more. In November 74% of AI sessions were engaged (organic: 61%), with 3X the key events per session. The assistant has already done the comparison, so the visitor comes to act.

AI visitors sign up far more oftenTwo tools, two definitions, same answer.
AI assistants: reach the sign-up step (GA4, 28 days)53.1%
Google organic: reach the sign-up step11%
AI search: visit to signup (Mixpanel, 6 months)35.6%
Google organic: visit to signup12.7%
Source: client GA4 funnel exploration and Mixpanel, anonymised
Result 4
86%
of AI-referred signups came from ChatGPT

ChatGPT sends most of the signups. Gemini is next.

ChatGPT brought 86% of AI search signups over six months and Gemini 10%. Claude, Perplexity and Copilot shared the rest.

That is why we track buyer prompts in both ChatGPT and Gemini, and why the pages we built answer the comparison and pricing questions people put to them.

Where AI-referred signups came fromShare of AI search signups by assistant, last 6 months.
86%10%4%
ChatGPT 86%Gemini 10%Claude, Perplexity and Copilot 4%
Source: client Mixpanel, anonymised
Result 5
2.9X
AI answers mentioning the brand, alongside 2.6X Google AI feature impressions

Visibility grew before the signups did

Google Search Console showed impressions in AI Overviews and AI Mode up 2.6X over 90 days. Ahrefs found 2.9X more AI answers mentioning the brand in its latest reading.

Visibility leads and traffic follows: AI exposure was rising across the whole quarter, and the signups caught up in November.

AI search visibility, before and afterIndexed. Search Console: last 90 days vs the previous 90. Ahrefs: latest reading vs the previous one.
Google AI Overviews + AI Mode impressions
100261
AI answers that mention the brand
100294
Before (= 100)After
Source: client Search Console (Generative AI features) and Ahrefs, anonymised
What we did

The GEO playbook, in six moves

No ads, no tricks. The same order we use for every GEO engagement.

  1. 01

    Measure AI properly first

    We built an AI Assistants channel in GA4 and a channel model in Mixpanel that separates AI search from paid ads, direct traffic and AI app integrations. The old dashboard had been counting paid ad clicks as organic and hiding ChatGPT visits inside Direct. We also started tracking 16 buyer prompts in ChatGPT and Gemini.

  2. 02

    Make the site readable to AI crawlers

    Key pages pointed their canonical tags at an app screen, and dozens of client-rendered pages served a near-empty shell to bots. We fixed canonicals, server-rendered the links that matter, added structured data and checked access for 22 AI and search crawler user agents.

  3. 03

    Rebuild content around the questions buyers ask AI

    We split overloaded posts into focused, answer-first pages: pricing, free plan, comparisons and use cases. We briefed 40+ use-case landing pages and refreshed 15 posts from Search Console data, each opening with a direct answer an assistant can quote.

  4. 04

    Publish the pages AI likes to cite

    Comparison and alternatives pages with clear verdicts, tables and FAQs. Assistants lean on these when someone asks which tool to choose.

  5. 05

    Fix internal linking

    A full link audit found only 12% of internal links sat in body copy, and a quarter of pages had no in-content links pointing to them. We delivered 1,600+ link fixes so authority flows to the pages that convert, and so crawlers find them.

  6. 06

    Earn mentions where AI looks

    Product listings, directories and roundup articles. Assistants recommend brands that other sites already recommend, so off-site mentions matter as much as on-site pages.

What it means for you

Four lessons from this GEO case study

1

Fix the measurement before you judge GEO

If AI visits land in Direct and paid clicks land in Organic, you cannot see what is working. Fixing it changed every decision that followed.

2

AI traffic is small but it converts

AI assistants were about 4% of sessions in November, yet those visitors signed up at almost three times the organic rate. Judge it on signups, not sessions.

3

Crawlability is the first GEO lever

An assistant cannot recommend a page it cannot read. Canonical bugs and client-side rendering hid key pages from AI crawlers for months.

4

Growth is uneven, then sudden

AI-referred signups were roughly flat for four months, then jumped once the technical fixes and the new pages were all live and recrawled.

Method and data notes

How we got these numbers

  • Mixpanel. Unique signups and visitors by channel, using a channel model Top Search Agency built in the client’s Mixpanel. AI search = visits whose UTM source or referrer is an AI assistant. AI app integrations and paid ads are separate channels. Pulled 16 November 2026.
  • GA4. Sessions, engaged sessions and key events by an AI Assistants channel group placed above organic search. GA4 explorations keep 2 months of data on this property, so the GA4 trend starts on 15 September.
  • Search Console and Ahrefs. AI Overviews and AI Mode impressions, last 90 days to 12 November vs the previous 90. Ahrefs AI responses mentioning the brand, reading of 15 November vs the previous one.
  • Anonymised. No client name, product category or volumes. Every figure is a ratio, a percentage or an index, and every chart is redrawn from source data.
  • Honest caveats. November is a part month, so we compare per-day rates or say so. AI assistants are still about 4% of sessions. Growth happened during our engagement alongside the client’s own product and marketing work.
FAQ

GEO case study: common questions

What is GEO (generative engine optimization)?

GEO is the work of getting your brand named, cited and linked in AI answers from ChatGPT, Gemini, Perplexity, Claude, Copilot and Google’s AI Overviews and AI Mode. It builds on SEO but adds AI crawler access, answer-first content and off-site mentions. Read our GEO service page for the full approach.

How did you measure AI-referred signups?

In Mixpanel we built a channel model that labels a visit as AI search when its source or referrer is an AI assistant, such as chatgpt.com or gemini.google.com, and keeps AI app integrations and paid ads out of it. In GA4 we added an AI Assistants channel above organic search. Both tools show the same trend.

Why is the client anonymised?

The client asked us not to publish its name or traffic volumes. We show multiples, percentages and indexed values instead, and every chart is redrawn from the source data.

How long did it take to see results?

The programme started in September 2026. AI-referred signups were flat through October and jumped in November, once the technical fixes and new pages were live and recrawled. Plan for one to three months before AI traffic moves, depending on how quickly fixes ship.

Is AI search worth it if it sends so few visitors?

Yes, if you judge it on conversions. Here AI assistants sent about 4% of sessions but their visitors signed up at 2.8X the rate of Google organic, and the share is growing every month.

Can you do the same for my company?

Start with a free AI SEO audit. We check whether AI crawlers can read your site, which buyer prompts name you or your rivals, and how your analytics count AI traffic.

Your turn

Want AI assistants to recommend you?

We will show you where you stand in ChatGPT, Gemini and Google’s AI answers today, and what to fix first.

©2026 Top Search Agency. Founded 2024 in Islamabad, Pakistan. All rights reserved.

Book a 30-min callGet my free audit
Scroll to Top