Search engines are no longer the only platforms shaping purchase decisions. ChatGPT handles over 100 million queries per week (OpenAI, 2024), Perplexity claims 10 million monthly active users, and Google AI Overviews now appear on a growing share of search results. When a business owner asks an AI assistant to recommend a service provider or compare solutions, your company is either mentioned or absent. GEO (Generative Engine Optimization) is the set of practices designed to influence these AI-generated responses.
What AI search engines change for business visibility
A traditional search engine displays a list of links. The user clicks, compares, goes back. AI-powered engines work differently: they synthesise information and deliver a single answer, sometimes with no outbound links at all. The user gets what they need without visiting your site.
This shift has direct consequences. According to an Authoritas study (2024), Google AI Overviews reduce organic click-through rates by 18 to 64% on queries where they appear. For SMEs that depend on organic traffic, this trend demands swift adaptation.
The good news: LLMs (Large Language Models, the technology behind ChatGPT and Perplexity) do not fabricate answers from nothing. They rely on indexed web sources, databases and structured content. Understanding how these models select their sources is the foundation of GEO.
How LLMs choose which sources to cite
ChatGPT, Perplexity and Gemini operate differently, but their source selection logic shares common ground.
LLMs favour content that demonstrates verifiable expertise. An article that cites data, references sources and clearly identifies its author has a stronger chance of being selected than a generic text without attribution. This criterion aligns with Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness), already used to evaluate page quality.
Content structure also plays a decisive role. Language models extract information in blocks: a paragraph that directly answers a question, a list of criteria, a concise definition. Well-segmented text with descriptive subheadings facilitates this extraction.
Perplexity operates as an AI-augmented search engine, indexing the web in real time and citing sources with links. ChatGPT draws on its training data and, in its web browsing version, on live search results. Google AI Overviews pull from Google's standard index and favour pages already ranking in the top positions.
Six pillars of an effective GEO strategy
Generative Engine Optimization rests on concrete principles you can apply today.
Structure content for extraction. Each H2 section should read as a standalone block. No references to "as mentioned above" or "the previous section." An LLM may extract any block without surrounding context.
Include sourced statistics. Language models favour claims backed by verifiable numbers. Mention the source, year and context for every statistic.
Define technical terms. When you use an acronym or specialised concept, add a concise definition at first mention. This increases the probability that the LLM retains your phrasing.
Build E-E-A-T signals. Detailed author pages, complete legal notices, verified customer reviews and displayed certifications. These credibility signals influence the trust AI models place in your content.
Implement schema markup. Structured data (FAQ, HowTo, LocalBusiness, Article) helps automated systems understand your content. A well-implemented FAQ schema increases the likelihood of appearing in AI Overviews.
Publish consistently. Content freshness weighs in source selection. A site updated monthly signals to crawlers and LLMs that the information is current.
Google AI Overviews: adapting your existing SEO
Google AI Overviews sit at the convergence of traditional SEO and GEO. They appear directly in search results as an AI-generated panel above organic listings.
To feature in them, your page must already rank in the top 10 organically. AI Overviews draw primarily from pages ranking between positions 1 and 10. Working on your traditional SEO therefore remains a prerequisite.
The difference from standard SEO lies in content format. Google AI Overviews extract short, factual passages. Paragraphs of 3 to 4 sentences that directly answer a specific question are favoured. Numbered lists and comparison tables are also overrepresented in AI panels.
A SEMrush study (2024) shows that 62% of sources cited in AI Overviews contain structured lists or tables. This format facilitates automatic information extraction.
Perplexity and ChatGPT: two distinct approaches
Perplexity works as a conversational search engine. It indexes the web continuously, formulates a synthesised answer and cites its sources with numbered links. To appear in Perplexity, the principles of standard SEO largely apply: quality content, solid indexation and domain authority.
What sets Perplexity apart is its preference for recent and factual content. A blog post published three years ago with outdated statistics will be bypassed in favour of updated content. The last-modified date, visible in the HTML or sitemap, influences selection.
ChatGPT with active web browsing functions differently. The model performs real-time searches via Bing, reviews results and synthesises the information. Presence in the Bing index, often overlooked by practitioners focused on Google, becomes a visibility factor.
For the version of ChatGPT without browsing, influence is exerted during model training. Content that is widely cited, shared and linked across the web has a higher chance of being included in training data. This logic further reinforces the value of link building and brand awareness.
The connection between GEO and your broader AI strategy
GEO does not operate in isolation. It fits within a wider approach to integrating artificial intelligence into a company's digital strategy.
Consider a concrete example. A physiotherapy practice in the French Alps publishes a detailed article on knee rehabilitation after ACL surgery. The article cites medical studies, defines anatomical terms, includes FAQ schema and clearly identifies the author (name, qualifications, experience). When a patient asks ChatGPT which practice to consult for post-operative rehabilitation near their location, that content stands a strong chance of being referenced.
The same logic applies to tradespeople, retailers and service providers. The content you publish today shapes how AI assistants will talk about your business tomorrow. This long-term perspective aligns with SEO, making the two disciplines naturally complementary.
Measuring your visibility in AI responses
Unlike traditional SEO where Search Console provides precise data, GEO measurement remains complex. No official dashboard aggregates your presence across ChatGPT or Perplexity responses.
Several approaches allow you to track visibility nonetheless. The first involves manually testing your target queries in each AI engine and documenting the results. It is manual work, but it proves instructive.
Specialised tools are starting to emerge. Otterly.ai and Peec.ai offer automated tracking of mentions in AI responses. These solutions are still young but provide a first view of your presence.
Monitoring referral traffic from AI domains serves as an indirect indicator. In Google Analytics 4, filter traffic by source to identify visits from chat.openai.com, perplexity.ai or other conversational platforms. Growth in this traffic signals that your content is being cited.
GEO action plan for an SME
Start by auditing your existing content. Identify pages that address your areas of expertise and evaluate their structure: are the H2s self-contained? Are data points sourced? Is the author identified? This audit typically reveals that most existing pages need restructuring rather than rewriting from scratch.
Prioritise rewriting high-potential pages next. Your service pages and most substantial blog articles are the primary targets. Add sourced statistics, structure content into extractable blocks and implement schema markup. Focus on pages that already rank well in organic search, since these are the most likely candidates for AI Overviews inclusion.
Set up a regular publishing schedule. One article per month, updated quarterly, is enough to signal to LLMs that your site is an active source. Each article should follow GEO best practices from creation rather than requiring optimisation later. Building these habits into your editorial workflow saves time and produces consistently structured output.
Work on your off-site presence. Mentions of your brand on other websites, professional directories, forums and social media feed the knowledge that LLMs have of your business. Link building takes on an additional dimension in this context. A brand mentioned across multiple authoritative sources carries more weight in LLM responses than a brand only referenced on its own website.
Finally, monitor results. Set a monthly review to test your core queries across ChatGPT, Perplexity and Google AI Overviews. Document which pages get cited and which do not. This feedback loop allows you to refine your approach over time, doubling down on content formats and topics that gain traction in AI-generated responses.
Frequently asked questions
Do you need to optimise for each AI engine separately?
The fundamentals are shared: structured content, sourced data, displayed expertise. Only a few specifics differ. Perplexity favours recent content, ChatGPT with browsing relies on Bing, and Google AI Overviews pull from the Google index. Working on overall content quality covers the majority of cases.
Does GEO replace SEO?
No. GEO complements SEO. Pages that rank well organically are also the ones LLMs tend to cite. Abandoning organic search in favour of GEO alone would be counterproductive.
How long does it take to appear in AI responses?
There is no guaranteed timeline. Updated and heavily linked content surfaces faster. Perplexity indexes in real time, so the effect can be near-immediate. For ChatGPT without browsing, the delay depends on model training cycles.
Is schema markup essential for GEO?
It is not mandatory but strongly recommended. Structured data facilitates automated understanding of your content. FAQ, Article or LocalBusiness schema increases the chances of extraction by AI systems.
Are local businesses affected by GEO?
Yes. AI assistants handle a growing volume of local queries. A user asking Perplexity for a reliable plumber near Annecy will receive an answer based on local web content. Businesses that have invested in structured, sourced content will be cited.
