Case studies
Real work. Data you can check.
We publish a result only when we can show the report behind it. Two clients, two markets, and an honest account of what each study does and does not prove.
No contract. We reply within one working day.
Two clients, two markets, real reports. These are not forecasts for your site.
We keep this section short on purpose. Two studies we can document in full are worth more to a buyer than twenty vague claims, and we would rather add a third when it is ready than pad the list.
The studies
Read the two studies
Each one follows the same order: the problem, the work, the measurement, the result and the limits.
Emtech, Dubai
A UAE IT company rebuilt for search and AI. We repaired crawler access, rebuilt service pages to answer buyer questions and added a working lead path.
Sudarshan Stoneix
A marble and granite company built from zero. Product pages, technical SEO and entity work in six months, measured in Search Console and AI mention tracking.
One client per market. Results depend on your starting point, so treat these as examples of method, not promises.
How we report
Where the numbers come from
Different numbers deserve different levels of trust, so we label the source next to each one. Google Search Console is first party data from Google, and we treat clicks and impressions from it as the firmest figures on a page.
AI mention counts are different. They come from third party visibility tools, and they depend on the tool, the prompt set and the date. AI answers also change from run to run, by user, place and time. That is why we show dated screenshots and describe patterns instead of promising a fixed count or position.
Screenshots are part of the evidence, not decoration. A screenshot proves what one engine said to one query at one moment. Repeated captures over time show whether a pattern holds. We say which is which.
Finally, we state what a study does not show. Several changes usually happen together, so credit cannot be split cleanly between them. One client in one market is not your market. We would rather say that on the page than leave you to find out later.
What to look for
How to judge any case study, ours included
Is the source named?
Search Console, an AI visibility tool, or a dated screenshot. If the source is missing, the number is a claim.
Is there a time window?
A figure without a period tells you little. Ours say six months or name the day the captures were taken.
Is there a baseline?
A result means more beside a starting point. The Sudarshan Stoneix study includes an earlier baseline of zero tracked mentions.
Are the limits stated?
Every honest study says what it cannot prove, such as one cause, another market or another query.
Can the client be named?
We name Emtech and Sudarshan Stoneix. We do not use anonymous logos or unnamed brands as proof.
Are promises avoided?
No study can guarantee a future ranking or citation. Be wary of any that implies it will.
Patterns
What the two studies have in common
Different trades, the same order of work.
Emtech sells IT contracts to businesses in the UAE. Sudarshan Stoneix sells marble and granite. The products could not be more different, but the first problem was similar: neither company was easy for a search engine or an AI assistant to read and understand.
In both cases we started with access and clean data, moved to pages that answer buyer questions directly, made the company easier to identify, and measured the result. The detail differs. The order does not.
That is the useful lesson for your site. The fix list is rarely exotic. It is usually a handful of basic problems that nobody has looked at in order.
- Fix access and foundation first
- Answer buyer questions on the page
- Make the company easy to identify
- Measure with named sources
Let crawlers read the site.
Say it plainly, early on the page.
Same facts everywhere.
Named source, dated, repeated.
Structure
What every case study page contains
- 1
The problem
What was stopping the company being found, understood or contacted, described in plain words.
- 2
The work
What we changed and in what order, with the reasoning, so you can judge whether the method would suit you.
- 3
The measurement
Which sources the numbers come from, over what period, and how much trust each deserves.
- 4
The result
The figures as reported, with screenshots of the original reports or captured answers.
- 5
The limits and hindsight
What the study cannot prove and what we would do differently next time.
A reading guide
How to read our case studies
A case study is a story with numbers in it, and stories are easy to read too generously. Here is a way to read ours, and anyone else's, with more care. Start with the question the study answers. Emtech is about being cited for IT AMC queries in AI engines. Sudarshan Stoneix is about growth from zero in Google search and in tracked AI mentions. They are different questions, so the numbers are not comparable.
Next, find the source of each figure. Search Console numbers come from Google and are the firmest on the page. AI mention counts come from a third party tool and depend on that tool's prompts and its date. Screenshots of AI answers show one answer, from one engine, at one moment. Our Emtech study and the Sudarshan Stoneix study label the source next to each result.
Then look at the time window and the starting point. A result after six months from a starting point of zero tells a different story from a result after six weeks on a long established site. If a study omits either, ask for it.
- Read the question first
- Match each number to its source
- Check the time window and the baseline
Compare
What each kind of evidence can and cannot show
| Evidence | What it can show | What it cannot show |
|---|---|---|
| Impressions, clicks and average position in Google over a stated period | Why the change happened, or what any AI engine said | |
| Third party AI mention counts | A pattern of brand mentions across tracked prompts | What a real buyer saw, or a fixed count you can expect |
| Screenshots of AI answers | What one engine said to one prompt on one day | That the answer is stable, or that every user sees it |
| Before and after prompt tests | Whether the answers moved on the tested prompt set | That the same prompts represent all your buyers |
| Written account of the work | Which changes were made and in what order | How much of the result each change caused |
No study on this page adds new numbers beyond those already shown on each case page. We prefer to show fewer figures that we can source.
The problem
The limits of any case study, including ours
Several things changed at once
A site rarely gets one fix. Access, content, structure and entity facts moved together, so credit cannot be split cleanly.
One client in one market
Emtech is a Dubai IT company. Sudarshan Stoneix is a marble and granite company. Your market, competitors and starting point differ.
AI answers drift
Engines update, and answers differ by user, place and time. A capture is a snapshot, not a permanent position.
Tools measure differently
Two AI mention trackers can give different counts for the same brand. The method matters as much as the number.
Survivorship
Published studies are the ones we can document in full. That is deliberate, but it does mean they are not an average of everything that could happen.
Using them
How to use a case study to make a decision
- 1
Find the nearest match
Which study is closer to your situation: a services company selling contracts in the UAE, or a product company growing from scratch? Read that one in full.
- 2
Take the method, not the number
The order of work is the useful part: access, answers, entity facts, then measurement. Ask whether those same problems exist on your site.
- 3
Check your starting point
Run the free AI crawler access checker and test your buyer questions. If you have the same basic problems, the method applies. If not, tell us what is different.
- 4
Ask what we would test
Ask us what prompts we would use and which engines we would check for your business. A clear answer is a good sign.
- 5
Decide on the scope
The GEO Pilot is a fixed $999, delivered in 30 days. It ends with a before and after test, so you see your own evidence rather than relying on ours.
What we do not publish
Things you will not find on this page
You will not find anonymous logos, unnamed brands, testimonials without a source or a count of clients we have not documented. You will not find a promise that you can expect the same result. You will not find a forecast of rankings, citations or traffic.
You will also not find numbers added here that are missing from the case pages. This page points to the two studies and explains how to read them. The figures live with their sources.
We add a new study only when we can show the report behind it and the client agrees to be named. That means the list may stay short for some time, and we think that is better than padding it.
Questions to bring
What to ask any agency about its results
Can I see the report?
Ask for the original report or a dated screenshot, not only a summary slide.
What period does it cover?
A figure without a start and end date is hard to judge.
What was the baseline?
A result means little without the starting point.
Is the client named?
Named clients can be checked. Anonymous ones cannot.
What does it not show?
An honest answer names the limits. A defensive answer is a warning sign.
Does it promise anything?
Be wary if a study implies you will get the same outcome.
FAQ
Case study questions
Where do the numbers come from?
Search Console figures come from Google. AI mention counts come from third party visibility tools. AI answer screenshots were captured from live searches on the day. We name the source next to each figure.
Why only two case studies?
We publish a result only when we can show the report behind it. We would rather show two documented studies than a longer list we cannot support.
Are the AI answers in the screenshots permanent?
No. AI answers change by user, place and time. We repeat checks and describe patterns, not a fixed position.
Will I get the same results?
We cannot promise that. Results depend on your starting point, your market and your competition. An audit shows what is in your way and what a sensible first step looks like.
Can I speak to these clients?
Ask us what we can share. We do not publish client contact details on this site, and we will tell you honestly what we are able to show you.
Do you publish dashboards or sample data?
Our product style visuals on this site use sample data and are labelled as such. The screenshots in the case studies are the real reports.
What will you show me in an audit?
Your buyer questions run through five AI engines, who is named, which sources are cited and a short ranked list of first fixes. It is free and there is no obligation.
Why do you show so few case studies?
We show a study only when we can document it in full and the client agrees to be named. Two documented studies are more useful to a buyer than a long list of claims we cannot support.
Can I trust AI mention counts?
Treat them as a pattern, not a fixed figure. Counts come from third party tools, and different tools can report different numbers. We always say which source a number comes from.
Are Search Console numbers more reliable than AI mention counts?
Search Console figures are first party data from Google, so we treat them as the firmest numbers on a page. They still show only Google search, not what any AI engine said.
Does one case study prove GEO works?
No. It shows what happened for one client in one market, and how we got there. It cannot prove cause, and it cannot tell you what will happen for you. The Pilot's retest gives you evidence on your own site.
Keep reading
Related pages
Emtech case study
Five AI engines, one query theme.
Read moreSudarshan Stoneix case study
Six months, measured in Search Console.
Read moreLLM visibility audit
Your free baseline in five engines.
Read moreGenerative engine optimization
How we earn citations.
Read moreServices and pricing
Published ranges.
Read moreKeep reading
Guides on this topic
How to get mentioned in ChatGPT: what to do, what to skip
No one can guarantee a ChatGPT mention, but you can remove blockers, make your facts clear and earn the sources it reads. A practical, honest checklist.
Read moreAI search traffic in Google Analytics: how to track ChatGPT and AI referrals in GA4
GA4 added an AI Assistant channel in May 2026, but AI Overviews clicks still sit in Organic Search and many AI visits arrive as Direct. Here is how to set up tracking, what volumes to expect and how to report it honestly.
Read moreAI visibility metrics: citation share, share of voice and mention rate defined
Mention rate, citation rate, citation share and AI share of voice are used loosely. Here is how to define and compute each from a fixed prompt set, what studies show about unstable answers, and how much a percentage can be trusted.
Read moreYour turn
Want a baseline for your brand?
Every engagement starts with an honest audit.
- Baseline recorded
- Ranked plan
- No obligation
WhatsApp: +91 63777 67206
Email: hello@thegrowingai.com







