GEO Pilot: we optimise your site for AI search in 30 days, $999.See what is included →

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.

Engines checked
  • Google AI Mode
  • ChatGPT
  • Gemini
  • Microsoft Copilot
  • Perplexity
5AI engines citing EmtechDubai IT company, AMC queries
1.48MGoogle impressions in six monthsSudarshan Stoneix, built from zero
909AI mentions trackedSame client, same period
17.1KGoogle clicks in six monthsSearch Console, same client

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.

UAE, IT services

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.

5AI engines citing Emtech
Manufacturing, export

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.

1.48MGoogle impressions in six months
909tracked AI mentions

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
1Access

Let crawlers read the site.

2Answers

Say it plainly, early on the page.

3Entity

Same facts everywhere.

4Measure

Named source, dated, repeated.

Structure

What every case study page contains

  1. 1

    The problem

    What was stopping the company being found, understood or contacted, described in plain words.

  2. 2

    The work

    What we changed and in what order, with the reasoning, so you can judge whether the method would suit you.

  3. 3

    The measurement

    Which sources the numbers come from, over what period, and how much trust each deserves.

  4. 4

    The result

    The figures as reported, with screenshots of the original reports or captured answers.

  5. 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

EvidenceWhat it can showWhat it cannot show
Search Console dataImpressions, clicks and average position in Google over a stated periodWhy the change happened, or what any AI engine said
Third party AI mention countsA pattern of brand mentions across tracked promptsWhat a real buyer saw, or a fixed count you can expect
Screenshots of AI answersWhat one engine said to one prompt on one dayThat the answer is stable, or that every user sees it
Before and after prompt testsWhether the answers moved on the tested prompt setThat the same prompts represent all your buyers
Written account of the workWhich changes were made and in what orderHow 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. 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. 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. 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. 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. 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.

Our clients

Businesses we work with

Real businesses we have worked with. We show their results only with the limits stated.

Quality Mascots Costumes logoRMS Stonex logoSudarshan Stoneix logo
Emtech Computers logo
Happy Home IC logo
Trade Like Master logo
One Percents logo
Arc Studio logo

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.

Your 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

No card, no contract. Nitesh replies himself, usually the same working day. By sending, you agree we may contact you about this request (privacy policy).