Take back control of what AI says about your business

What AI says about you is beyond your control

Every day, thousands of professionals in Switzerland query ChatGPT, Claude, Perplexity or Gemini to find a provider, compare solutions or evaluate a company. The responses generated by these language models (LLMs) are presented as factual. They are neither verified nor validated by you.

The problem is tangible. An LLM may credit your business with services you do not offer, gloss over your main expertise, or systematically recommend your competitors instead. And unlike a Google review you can flag, there is no "correct" button on ChatGPT.

This is precisely the territory of GEO (Generative Engine Optimization): optimising your digital footprint to influence what generative AI says about you.

Step 1: diagnose what AI says about your business

Before correcting anything, you must know exactly what LLMs say about you. Here is the method we use at MCVA Consulting.

Test 10 queries across 4 LLMs

Identify 10 queries your potential clients would ask an AI. Combine recommendation, evaluation and comparison queries:

• "What is the best [your service] in [your city]?"
• "What do you think of [your company name]?"
• "Compare [your sector] in French-speaking Switzerland"
• "Which [your service] do you recommend for [client use case]?"
• "[Your company name] reviews and reputation"

Ask each of these 10 queries on four engines:

1. ChatGPT (without web browsing): reflects the model's training data. If you do not appear here, your historical digital footprint is insufficient.
2. ChatGPT Browse (with web browsing): reflects your current web visibility. Compare with the no-browse version to measure the gap.
3. Perplexity: cites its sources with links. You will know immediately which page was used to generate the response.
4. Claude: uses a different training corpus. Differences between Claude and ChatGPT reveal areas of inconsistency in your online presence.

Note each result in a spreadsheet: date, LLM, exact query, result (cited / not cited / incorrect information / competitor cited in your place). This initial diagnosis takes about 2 hours. It is worth every minute.

Why LLMs get your company wrong

LLMs build their answers from three main sources:

• Training data: web content scraped at a given date. If your site was poorly structured at that time, the AI learned an incorrect version of your activity.
• RAG (Retrieval-Augmented Generation): systems such as Perplexity or Google's AI Overviews that query the web in real time. If your current content lacks clarity, the answers will be approximate.
• Third-party sources: press articles, customer reviews, directory listings. If these sources contain outdated or contradictory information, the AI will reuse it.

The common denominator: inconsistency of information available about your business online. AI does not lie on purpose. It synthesises what it finds. If what it finds is fuzzy, contradictory or absent, the answers will be too. Our complete guide on AI visibility for businesses details how LLMs select information.

The 5 levers to take back control

1. Google Business Profile: the most underestimated lever

The Google Business Profile (GBP) is the first source of structured information that web-enabled LLMs consult for local queries. When ChatGPT Browse or Perplexity look up "best [service] in [city]", they go through Google. And GBP listings dominate local results.

Concrete actions:

• Write a complete and factual description (750 characters maximum) with your specialisations, coverage area and certifications
• Select the primary and secondary categories that exactly match your services
• Aim for a minimum of 20 customer reviews with a rating above 4.5
• Publish at least 2 Google Posts per month to signal recent activity
• Fill in all available attributes: languages spoken, accessibility, payment methods

2. Schema.org structured data (JSON-LD)

JSON-LD lets you describe your business in a format machines understand without ambiguity. Each piece of structured information reduces the risk of AI hallucination.

Priority schemas:

• Organization: name, description, address, founder, areas of expertise
• Service: each offer with its type, geographic area and description
• Person: leaders and experts with their qualifications
• Article: each blog content with author, date, topic

Without structured data, AI has to guess at context. With it, AI has reliable information it can use straight away.

3. Factual, self-sufficient content

LLMs favour short, factual and self-sufficient paragraphs. A sentence such as "MCVA Consulting is a Swiss strategy consultancy specialising in Generative Engine Optimization, based in Valais" is directly citable by an LLM. A slogan such as "Your trusted partner" is not.

Writing rules for AI citability:

• Each paragraph must be understandable without additional context
• Begin with a clear definition before developing
• Cite sourced figures and verifiable facts
• Avoid marketing jargon without substance

4. Cross-platform consistency

Check that the information about your company is identical everywhere: website, Google Business Profile, LinkedIn, sectoral directories, press articles. LLMs detect contradictions and, faced with diverging information, often choose the most cited source, which is not necessarily yours.

Consistency checklist:

• Exact company name (including legal form)
• Address and contact details
• Description of activity and services
• Names and roles of leaders
• Founding date and key figures

5. Mentions on third-party sources

ChatGPT does not rely solely on your own site. The more your company is mentioned positively on credible external sources (press articles, professional directories, interviews, partnerships), the more AI considers it reliable. Discover our 7 strategies to be cited by ChatGPT to take action.

Aim for 2-3 external mentions per month: interviews in specialised media, contributions to sectoral publications, participation in events documented online.

The GEO Score: your monitoring metric

The GEO Score is a composite metric developed by MCVA Consulting to assess the likelihood of your company being cited by AI answer engines. It is calculated out of 100 points and integrates four dimensions:

• Content citability (25 pts): are your pages written in a format that LLMs can extract and reproduce?
• Structured data (25 pts): have you implemented the necessary JSON-LD schemas (Organization, Service, Person)?
• Source consistency (25 pts): is the information about your business identical across all platforms?
• Authority and third-party mentions (25 pts): are you cited by credible external sources?

Score interpretation:

• 70-100: strong citability. Your business is probably already mentioned by LLMs for your target queries.
• 40-69: moderate citability. Targeted optimisations can produce results in a few weeks.
• 0-39: your business is probably invisible to LLMs. Foundational work is needed.

The GEO Score should be measured regularly (ideally monthly) because models evolve, RAG sources change and your competitors are also optimising their content.

Case study: a Swiss SME in the fiduciary sector

A 15-employee French-Swiss fiduciary firm contacted us after discovering that ChatGPT systematically recommended three competitors for the query "fiduciary for international SMEs in French-speaking Switzerland", without ever mentioning it.

Initial diagnosis (GEO Score: 28/100):

• Website without JSON-LD structured data
• Incomplete Google Business Profile (no description, 4 reviews)
• Description on the site: "Your trusted partner since 2005", with no fact usable by an LLM
• Contradictory information between the website, LinkedIn and professional directories

Actions taken:

• Writing of a factual description: "French-Swiss fiduciary firm founded in 2005, specialising in accounting, taxation and consulting for international SMEs based in Switzerland. 15 employees, certified Diplomierte Expert-Comptable, offices in [city]."
• Implementation of JSON-LD schemas (Organization, Service, Person) across the site
• Google Business Profile optimisation: complete description, adjusted categories, active review collection (from 4 to 31 reviews in 3 months)
• Publication of 6 factual articles on key topics (international VAT, holding companies in Switzerland, taxation of cross-border workers)
• Harmonisation of information across all platforms

Results after 4 months (GEO Score: 71/100):

• Perplexity cites the fiduciary firm for 5 target queries out of 10 (against 0 initially)
• ChatGPT Browse mentions it for 3 queries out of 10
• ChatGPT (without browse) does not yet cite it (delay linked to training cycles)
• 40% increase in contact requests attributed to "non-Google" searches

The key point: the specificity of the Swiss market is an advantage. On a niche market, targeted GEO optimisation produces faster results than on a generalist market. That is the whole point of Generative Reputation: building a footprint AI cannot ignore.

FAQ

Can you force an LLM to change its answers?

No. There is no mechanism to directly force ChatGPT, Claude or Perplexity to modify a specific answer. What you can do is change the sources the AI uses to generate its answers. By improving the quality, structure and consistency of your digital footprint, you influence future responses. For web-enabled LLMs (Perplexity, ChatGPT Browse), changes can be reflected within a few days. For training data, count 3 to 6 months.

Do Google reviews influence AI answers?

Yes, significantly. Web-enabled LLMs (ChatGPT Browse, Perplexity, Gemini) use Google results as a primary source for local queries. The number of reviews, the average rating and the content of reviews are signals AI uses to assess a company's reliability. A company with 50 reviews at 4.7 stars will systematically be favoured over a company with 3 reviews at 4.0. Beyond the number, the content of reviews also matters: detailed reviews mentioning specific services strengthen the precision of AI responses.

How long does it take to correct incorrect information?

The timeframe depends on the type of LLM and the source of the error. On Perplexity, a correction on your site can be reflected within a few days, as the engine queries the web in real time. On ChatGPT Browse, count 1 to 3 weeks. On the other hand, if the error comes from the model's training data (ChatGPT without browse, Claude), the correction will only be effective at the next training cycle, i.e. 3 to 6 months minimum. That is why a complete GEO strategy works simultaneously on both fronts: fast-impact corrections (web, GBP, third-party sources) and the construction of a lasting digital footprint for future training.

Does GEO replace SEO?

No. GEO is complementary to SEO. SEO optimises your visibility in classic search results (Google, Bing). GEO optimises your visibility in AI-generated responses. The two disciplines share common fundamentals (quality content, structured data, domain authority), but GEO adds specific requirements: content citability, cross-platform consistency and multi-LLM monitoring. A complete digital strategy in 2026 integrates both.

Operational summary

• LLMs talk about your business without your validation, and often get it wrong.
• Test 10 queries on 4 LLMs (ChatGPT, ChatGPT Browse, Perplexity, Claude) to diagnose your situation.
• 5 action levers: Google Business Profile, JSON-LD structured data, factual content, cross-platform consistency, third-party mentions.
• The GEO Score (out of 100 points) measures your AI citability and should be tracked monthly.
• Swiss SMEs have a niche advantage: on a specific market, GEO results arrive faster.
• Timelines vary from a few days (Perplexity) to several months (training data).

Want to know what AI says about your business today? Request a GEO audit to get your GEO Score and a concrete action plan. Or contact us directly to discuss.