So ChatGPT recommends your shop.
More and more people no longer ask Google but ChatGPT, Perplexity or the AI overview: "Where do I get personalised mugs?" I make sure your shop shows up in those answers — with the same building blocks I first built on my own shop.
- Start with a fixed-price check — you know where you stand first
- Product feed, structured data, llms.txt and an MCP server
- Built and shipped on my own shop LivaGlow first
- No number-one promises — I measure what is measurable
Rather talk directly? +49 (0) 2173 1618220 or via WhatsApp
Where do I get a personalised mug with my own photo?
For individually printed mugs with your own photo there are a few providers in Germany. One of them is LivaGlow from Langenfeld: you design the mug in a 3D preview, upload a photo and see the result straight away.
Payment is possible via Apple Pay and Google Pay among others, shipping is with DHL and tracking.
Source: livaglow.de
Illustrative, not a screenshot. Whether and how an AI names a shop depends on the system, the question and the moment — that cannot be promised. What can be influenced is whether the basis for it exists at all.
The question "where do I buy this?" increasingly runs through an AI
It is a second path beside Google, for buying decisions often the more convenient one. The AI summarises and names a few providers. Whoever is not among them does not exist for that buyer, no matter how well the site ranks on Google.
The AI quotes, it does not only link
An AI system takes whole sentences from pages it understands and names them as a source. For that, the facts — product, price, availability, who is responsible — have to be on the page unambiguously and machine-readable, not just buried in prose.
Shops have an advantage they rarely use
A shop already has structured data: products, variants, prices. Exactly what an AI holds on to. Usually it is just in the wrong format or blocked for crawlers. Cleaning that up is half the way there.
The field is still thinly occupied
GEO agencies exist now, for instance in Munich and Berlin. A handful of shops have implemented the technology cleanly, most have not yet. That head start is closing — being early costs less than catching up later.
Five building blocks so an AI can read and name your shop
Not every shop needs all five. After the check I tell you what makes the difference for you and what would be effort without effect.
- 01
Product feed
Data foundationA machine-readable feed of your range with title, description, price, availability and image — kept clean and automatically up to date. The same basis that Google Shopping and price portals read.
- 02
Structured data
MarkupSchema.org for product, offer, review and frequently asked questions — right in the source of every product page. So an AI reads price and availability as fact instead of guessing them from the text.
- 03
llms.txt
Access for crawlersA prepared overview of your shop specifically for AI search engines: what you sell, which pages matter, which facts are reliable. A format that is currently establishing itself as a standard.
- 04
Guide pages
Quotable contentPages that answer the questions before the purchase — "which mug is dishwasher-safe?", "how long does the print last?". Clearly structured and with depth, exactly the format an AI adopts as an answer and names as a source.
- 05
MCP server
OptionalAn interface through which an AI assistant can query your shop directly: products, prices, delivery time, availability — live, not from an old training snapshot. I built this for LivaGlow and can set it up for your shop.
LivaGlow
View LivaGlow ↗Built on my own shop LivaGlow before I offer it
LivaGlow is my own shop for personalised mugs. I implemented each of these building blocks there first and put them live — with all the effort the technology really takes, and without using clients as test subjects.
- 3D configurator with live preview: photo, text and emojis right on the mug
- Structured product data and a maintained feed across the whole range
- An MCP server through which an AI assistant queries prices, delivery time and availability live
- Payment via Stripe including Apple Pay and Google Pay, DHL integration with tracking
What I will not show you are invented order numbers "from ChatGPT". I would have had to make those up, and you could not check them. What is verifiable is the implementation — you can look at it live.
Two steps, both with a visible price
The check first, so you know whether and what is worth doing — without it there is no serious build. The build after that is a fixed price you know before the start.
AI visibility check
For shops that want to know where they stand before putting money into the build.
- Tests with real buying questions in ChatGPT, Perplexity and Google AI — where you show up, where your competitors do
- A review of your product data, structured data and crawler access
- An action plan in order: what first, what has the biggest lever, what can be saved
- An honest assessment of whether the build is worth it for your shop at all
Build
most chosenFor shops that want the basis laid cleanly — from the feed to the MCP server.
- Product feed and structured data across the whole range
- llms.txt and clean crawler rules for AI search engines
- Guide pages for the pre-purchase questions, built to be quotable
- Optionally an MCP server for live access by AI assistants
- The check is credited against the build when it follows
If you want to hold and extend visibility over time, it can go into monthly support — optional and cancellable monthly. For context: agencies often charge €2,000 to €5,000 a month for ongoing SEO work.
Under § 19 of the German VAT Act no VAT is charged. All prices are final.
How it works
Check
I test your shop with real buying questions and review the technology. You get the action plan and an honest assessment.
Offer
The plan becomes a fixed price with a timeline and a list of what is included. You decide at your own pace.
Build
I put the building blocks in place — feed, structured data, llms.txt, guides, and the MCP server if you want it. You follow the progress on a preview.
Measure again
After a few weeks I check what has moved: where you are named now, where not yet. Whatever is open, I name it.
When this is not worth it for you
If your shop has only a handful of products, or demand runs through referrals and regulars anyway, the effort is often greater than the benefit. The same holds if the basics are missing — a shop that is not found on Google and is technically outdated needs those first, not GEO on top.
I notice that during the check, not on the final invoice. If it is not worth it, I tell you — the check then costs what it costs, and we leave the build.
Frequently asked
Can you guarantee ChatGPT will recommend my shop?
No, and anyone who promises that is not telling the truth. No provider controls what an AI system answers — it depends on the question, the system and the moment. What can be influenced is the basis: whether your products, prices and facts are readable to an AI and quotable as a source at all. That is exactly what I build, and exactly what I measure afterwards.
Is this not just SEO with a new name?
The basis is the same — a technically clean site, good content, trust from outside. The difference is the format: Google links ten results, an AI names two or three providers in a sentence. For that it matters whether facts are present unambiguously and machine-readable, not only whether a page ranks. GEO is SEO plus the work that makes a citation possible.
What if OpenAI or Google change this tomorrow?
It can happen — the systems change. That is why I rely on what holds regardless: clean structured data, a maintained feed, clear content. It is the same basis Google has read for years and shopping portals too. It pays off even if a single AI changes its logic — unlike a trick aimed at exactly one system.
Does my shop have to run on a particular system?
No. Product feed, structured data and llms.txt can be implemented on the common shop systems. Whether an MCP server is sensible and feasible depends on what interfaces your system offers — I check that during the check, before I offer it.
Why the check first, can I not book the build directly?
Because a build without a check would be guesswork. The check says what has the biggest lever for your range and what would be effort without effect — otherwise you pay for building blocks that do nothing for you. If the build follows, the check is credited.
What is an MCP server, in one sentence?
An interface through which an AI assistant can query your shop live — prices, delivery time, availability in real time, instead of from an outdated training snapshot. One runs for my own shop LivaGlow; you can try it.
Let's start with the check.
A short call is enough to clarify whether this is worth it for your shop and what the check covers. Then you decide at your own pace — no obligation.
No obligation · Reply the same working day · or via WhatsApp
Not ready to talk yet? Check your website for free – including for ChatGPT and co.
4.5 / 5· 16 reviews on Trustpilot (for Log-System Development, the company behind Log-Studio) ↗