Skip to main content
GEO optimisation
5 min readPublished on May 24, 2026Updated on September 09, 2026

ChatGPT Optimisation Services: What Is Actually Possible

Simon Heistermann

Simon Heistermann

Owner

This article was written with AI assistance and editorially reviewed.

Anyone who books a provider on the promise that ChatGPT will start recommending their business is buying a promise about a system that provider does not control. That is not a footnote, it is the core of any honest assessment of this service.

In short

ChatGPT optimisation is checkable technical work on crawler access, facts and structure - not a guarantee about what the model ends up answering. Understanding how ChatGPT actually gets its information makes it easy to see where a provider's work ends and the model's own decision begins.

How ChatGPT gets information about a business in the first place

ChatGPT draws on two separate sources. The first is the model's training data - a snapshot frozen at a point in time that no single optimisation effort can retroactively change. The second is web search, which ChatGPT uses for current queries when the feature is switched on: the system retrieves live pages, much like a classic search engine, and can cite from them.

For a service provider, that distinction is decisive. Work on your current website - structure, facts, crawler access - can directly influence the second source, once web search kicks in. Nobody has influence over the first source, the training data, because it only updates with a new model training run, and the selection process behind that is not public. A provider who doesn't volunteer this distinction has either not understood the mechanism or is deliberately glossing over it, because an honest explanation makes the offer sound less spectacular.

What a provider can actually influence in that process

The effective part of the work concentrates on the web-search source, because it's the only lever that can be worked on honestly:

  • Open robots.txt for the relevant AI crawlers, often blocked without anyone intending it
  • Build a clear business entity: who you are, what you offer, where, for whom
  • Date facts and phrase them verifiably instead of wrapping them in marketing language
  • Structure FAQ content so a single answer stays citable in isolation
  • Establish consistency between the website, Google Business Profile and other sources

Every one of these steps is checkable: you can look yourself at whether GPTBot is allowed in your robots.txt, whether your FAQ answers stand on their own, whether your details match across sources. A provider who names these steps concretely and shows you the result, instead of hiding behind "proprietary methodology," is working honestly.

Why a promise of a guaranteed mention does not hold up

Even with every one of these steps cleanly implemented, whether ChatGPT actually answers a given user question by naming your business remains open. The model selects sources by its own undisclosed criteria, weighs the exact phrasing of the user's question differently each time, and shifts its behaviour with every model update, with no advance notice to anyone outside OpenAI. A provider cannot get around that process, no matter how technically skilled the work is.

Anyone who still promises a fixed mention, a timeline for one, or a specific position in the answer is selling something outside anyone's control. That is the one sentence in this article that actually matters: ask every provider directly how they handle that limit. A provider who dodges the question, or offers a guarantee anyway, has either not understood the mechanism or is knowingly selling an empty promise.

How this differs from Perplexity and other systems

SystemShows sources visiblyWhat that means for the work
ChatGPTOnly with web search activeGroundwork mainly acts once search kicks in, not on raw training knowledge
PerplexityIn practically every answerCitable, clearly structured content shows a more direct effect here
Google AI OverviewsOften, with a linkOverlaps heavily with classic technical SEO

A strategy aimed exclusively at ChatGPT therefore covers only part of the relevant systems. The technical foundation - crawler access, verifiable facts, clean structure - largely carries over to all three, which keeps the extra effort for broader coverage manageable.

Concrete steps for the next 90 days

  • Days 1-30: check robots.txt and llms.txt, open them for the relevant AI crawlers, test the current state in ChatGPT with and without web search
  • Days 31-60: build the business entity and FAQ content, establish consistency between the website and external profiles
  • Days 61-90: document samples using your most important customer questions and repeat monthly

Conclusion

ChatGPT optimisation is solid technical work with a clear boundary: it raises the probability of a mention, it never guarantees one. Understanding that boundary lets you judge a provider's offer realistically instead of buying promises nobody can keep. How ChatGPT, Claude and Gemini differ in detail is in our article ChatGPT, Claude and Gemini Compared; how to systematically compare several providers is in Comparing AI Visibility Providers. For an informal assessment of your current visibility, feel free to get in touch.

Want to know if ChatGPT can even read your website today?

Get in touch

Frequently asked questions

Simon Heistermann

Simon Heistermann

Owner

Heistermann Solutions is the web studio run by Simon Heistermann. We build custom websites for small and medium-sized businesses that want to achieve more online.

Every article grows out of day-to-day project work and is reviewed editorially before publication.

Get it for free

Enter your email address. You'll immediately receive a confirmation link - after clicking it the checklist is available right away.

Let's talk about your project

Free introductory call