GEO: 5 powerful tips to boost your visibility with LLMs

/in /by

Good news: adapting to GEO (Generative Engine Optimization) doesn’t require relearning everything from scratch. It’s mainly about extending your solid SEO practices and developing new reflexes centered on structuring information and ensuring reliability. Here are the concrete areas to focus on to align your strategy with GEO requirements.

1. Create content designed for both AI and users

The question often arises: “Should we now write to please AI or to please humans?” The answer is simple: both at the same time. Fortunately, what’s good for the user (clarity, relevance) is also good for AI. In practice, revisit your key content with a fresh eye: if a paragraph were extracted on its own, would it still make complete sense? Does each section deliver a well-defined key idea?

  • One idea per paragraph: structure your text so that each paragraph conveys a single main idea. This creates clear units of meaning that AI can extract without requiring the full context. For example, avoid long, sprawling paragraphs that mix several concepts.

  • Explicit headings for each section: use subheadings (H2, H3, etc.) that clearly describe the content of the section that follows. These subheadings act as signposts for both your readers and the algorithm. A well-crafted heading helps AI understand, “Ah, this section is about the advantages of this solution,” and extract it accordingly.

  • Clear and natural language: avoid excessive jargon or convoluted phrasing. Prefer short, precise sentences, as if you were explaining to a smart beginner. An AI (just like a busy human reader) doesn’t have time to decipher complicated sentences. Tip: why not use tools (e.g., Hemingway or others) to flag overly long or confusing sentences in your drafts? This helps you simplify without dumbing down. In fact, shortening and clarifying your sentences makes life easier for AI.

  • Comprehensive coverage and expertise: GEO-friendly content is, above all, quality content. Keep covering your topics in depth, adding real value. An article that tackles a subject thoroughly and expertly has a greater chance of becoming the reference that AI cites. Don’t hesitate to include expert contributions, concrete examples, or case studies—these reinforce your content’s E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness). Show that you master your topic better than anyone else.

In short: keep producing excellent content as you would for SEO, but with greater emphasis on readability and precision.

2. Structure your pages to be “scannable” by AI

Structure is the backbone of GEO. Think of AI as a super-reader that skims your pages to extract the essentials without reading every single word. Make its job as easy as possible.

  • Leverage lists and easy-to-extract formats: generative answers often present information in bullet points, numbered steps, or comparison tables for clarity. If your content already contains these well-structured formats, you increase the chances of having parts reused as-is. For example, a well-crafted FAQ with a clearly posed question and a concise answer has a high chance of being picked up by AI to respond to the same query. Likewise, a clear comparison table (e.g., a “SEO vs GEO” table with columns) is ready-made for an AI wanting to explain differences. The same goes for standalone definitions (“Machine learning is…”) that a model can cite directly when someone asks for a definition. Pro tip: at the end of your sections or articles, include a short bullet-point summary of the key ideas—these recaps are a goldmine for generative models.

  • Use semantic HTML tags: make sure to use heading tags (H1, H2, H3, etc.) and list tags (<ul>, <ol> for bullets/numbers, <dl> for definition lists, etc.) properly in your HTML code. A well-formed hierarchy not only helps human readers navigate but also allows AI to better segment and contextualize each part. On the flip side, avoid a jumble of <div> without structure or headings out of order—that can confuse the algorithm. In practice: one <h2> per major section, <h3> for subsections, and if possible, use <section> or <article> elements in your code to clearly delimit logical content blocks. Each “chunk” of your page should be understandable on its own, and structure helps achieve that.

  • Keep paragraphs short and focused: as mentioned earlier, avoid overwhelming walls of text. Aim for 2–4 sentences per paragraph. A short paragraph should capture one clear unit of meaning that’s easy for AI to digest. Bonus: this also makes your content more pleasant for human readers.

  • Use visual snippets: whenever possible, enrich your pages with informative visuals—infographics, diagrams, images with explanatory captions. This not only engages human readers but also helps AI, which can pick up on captions and alt text to extract information. For example, a chart with a caption describing a key statistic can be referenced by AI (since generative engines also read captions and alt text).

In short: think UX + structure. What’s skimmable for a human is equally skimmable for AI. Well-structured content is “attractive” content for generative engines. And the good news? It’s exactly the same best practice we’ve been advocating for impatient internet users for years!

3. Focus on reliability: sources, citations, and updates

Trust is the new gold standard in GEO. Remember: an AI doesn’t “verify” the absolute truth of what you write, but it relies on content it deems trustworthy to build its answers. To earn that trust, your content must radiate credibility.

Here are some levers to activate:

  • Cite your sources explicitly: don’t just write “According to a recent study…” without further detail. Instead, say “According to a 2023 study by INSEE…”. Provide the author or organization, the date, and even the publication if relevant. Why? Because AI will recognize that the information is backed by a reliable source rather than a vague claim. By citing properly, you strengthen your informational legitimacy while giving AI useful context to assess credibility. Think about it: citing also means being read (by the machine). Content with clear references will be perceived as more trustworthy.

  • Include concrete data and figures: quantitative claims, statistics, and precise facts reinforce credibility. On top of that, these elements can be directly reused by AI in its answers. Example: “85% of consumers… (Source: Study X).” Sentences like this are highly likely to show up in a generated response to support a point—and you’ll be cited as the reference. One study even showed that including relevant data points can significantly increase content visibility in AI responses (up to +40%).

  • Keep your strategic content updated: AIs favor fresh and relevant information. While a model like GPT-4 may have a fixed knowledge cutoff, many generative engines (Bing, Google SGE, etc.) rely on real-time web data. Outdated content has less chance of being reused—or worse, it could cause the AI to spread inaccuracies. Identify your key pages (those that generate leads or answer common questions) and keep them fresh: updated numbers, recent examples, mentions of the latest trends. This ensures that when AI “picks up” your content, it finds valid, timely information. Google already values freshness in traditional SEO, and it’s very likely that AIs also prefer up-to-date content—especially for things like technology advice or market data.

By strengthening the reliability of your content, you win on two fronts: human readers will trust you, and AIs will view you as a credible source worth citing. Remember: in the AI ecosystem, perceived truth often prevails—and that perception largely comes from the quality of your sources and the seriousness of your writing.

4. Optimize the technical side: your site must be readable by AI

You can have the best content in the world, but if it isn’t technically accessible to search engines (classic or AI), it won’t matter. Technical optimization for GEO largely builds on the basics of technical SEO, with a few extra points to pay close attention to.

  • Ensure your pages are indexable by bots: this may sound obvious, but check that you haven’t accidentally blocked AI crawlers. They generally follow the same robots.txt rules as other bots. So, avoid inappropriate disallows. If, for some reason, you want to prevent your content from being used by AI (some companies do this for protected material), you can block them via robots.txt. But if your goal is maximum visibility, open the gates. In practice: no noindex tags on your key pages, no unnecessary restrictions for user-agents, etc. A page that isn’t crawled has zero chance of being cited.

  • Watch out for JavaScript-heavy content: this is a technical but crucial point. Current generative engines aren’t as good as Google at processing dynamically generated JavaScript content. Google has its rendering service (Google WRS) that executes JS to index what appears after page load. Most AIs don’t (yet) have such a full rendering service. As a result, if your site displays text only after a script runs (for example, client-side API-loaded content, or a 100% React app without SSR), the AI may see nothing at all. Make sure your essential content is present in the initial HTML or loads very quickly. If necessary, implement Server-Side Rendering (SSR) or dynamic rendering for important public pages, so even a basic crawler can access the content immediately. This is already a good SEO practice—and it’s even more important for GEO.

  • Optimize loading speed: a page that takes too long to load or requires 50 third-party scripts is bad for both users and bots. While there’s no formal proof that AIs prefer faster pages, it’s very likely (they aim to deliver quick answers “at record speed”). So, improve your Core Web Vitals scores, compress images, and remove unnecessary scripts. Provide fast access to information. At the very least, even if it doesn’t directly help the AI, your human visitors will thank you—and Google already takes that into account.

  • Leverage structured data (Schema): structured data (Schema.org) is a language used to “talk” to search engines. Google has been using it for years to deliver rich results. It’s highly likely that generative engines also rely on this markup to better understand elements like products, FAQs, reviews, etc. So keep integrating (or start integrating) relevant Schema markup on your key pages. Examples: FAQ schema on FAQ pages (to explicitly mark questions and answers), HowTo schema for tutorials, Product schema on product pages with attributes, etc. This markup isn’t visible on screen, but it provides extra context to bots about your page content. If it already helps get rich snippets on Google, it may well help tomorrow with rich answers in AI.

In summary: your technical foundation must be solid—a well-built, fast, easily crawlable, and parsable site. This is a prerequisite for SEO and just as crucial for GEO. What AI cannot see, it cannot cite—always keep that in mind.

5. Grow your “web footprint” and authority beyond your site

The last pillar of GEO—and by no means the least: to be the answer, you first need to be on the AI’s radar. A conversational AI doesn’t rely only on your official website. It learns and draws knowledge from a wide range of sources: Wikipedia, forums, online media, niche blogs, social networks, and more. If your name or content appears “nowhere” outside of your site, your chances of being cited by AI drop dramatically. That’s why it’s crucial to build your broader online presence—your digital footprint.

Here are a few ways to get there:

  • Contribute on other influential platforms: think about guest blogging on websites in your field, writing opinion pieces or guides for online media, publishing on LinkedIn if that’s where your industry is active, or answering questions on Quora/Reddit. The goal is to multiply the occasions where your expertise is visible outside your own site. Not so much for the backlink (though that can bring traffic), but so that your name or brand is regularly cited in relevant contexts. AIs pick up on these weak signals: if your company is frequently mentioned on a given topic, across different sources, they’ll identify you as part of the landscape on that subject.

  • Get mentioned in reliable sources: ideally, aim to be cited in studies, reports, interviews, podcasts, or conferences. If a government site or a reference report mentions your work, that’s powerful. But even short of that, being cited in industry newsletters, content aggregators, or articles by influencers in your niche contributes to your legitimacy. Traditional link-building sought backlinks; GEO values mentions themselves, even without links. The goal is to make your entity (person or brand) exist within the wider informational ecosystem.

  • Engage in communities: on specialized forums, Reddit, Stack Overflow (if relevant), or social networks, join the discussions in your field. By helping users and sharing valuable insights, you create content outside your site that models can ingest. For example, a detailed answer you post on a technical forum—even if it doesn’t bring direct leads—might one day serve as a source for an AI. At the same time, it reinforces your expert image among humans, often leading to more natural mentions of your name (a snowball effect for visibility).

  • Monitor your online reputation: keep an ear out for what’s being said about you. If misinformation circulates, correct it. For instance, an inaccurate Wikipedia article about your company, or a forum where someone shares outdated information about your services. Since AIs can repeat things they’ve read here or there, you’ll want to prevent errors from spreading. A concrete example: if ChatGPT incorrectly states that your product lacks a feature it actually has, it might be because of an old blog comment saying so; correct the source if possible.

By strengthening your presence beyond your site, you anchor your brand in the “memory” of the web. This way, AIs will have more material to draw from when answering questions in your domain. It’s long-term work that aligns with the principles of content marketing and personal branding—but in the GEO era, it takes on an even more strategic dimension.

Opportunities offered by GEO to boost your leads and competitiveness

Adopting GEO isn’t just about defensive prevention in the face of change—it’s also about seizing new growth opportunities for your business.

The first obvious opportunity: gaining visibility where your competitors aren’t yet present. If you start optimizing your content for AI now, you can occupy the space and become the go-to reference models prefer to cite. Being an early adopter of GEO gives you a lasting advantage, because once AI “gets used to” including you as a trusted source, it will be hard for a newcomer to take that spot. It’s a bit like being the first to rank on an SEO keyword 10 years ago—you benefit from a long head start.

Next, GEO can help you improve lead quality. As mentioned earlier, prospects coming through AI recommendations often arrive with stronger trust and clearer intent. They’ve asked the AI for advice or a product, and your offer was suggested—immediately placing you at the top of their list. The conversion rate of these leads may be higher than that of generic SEO traffic. Admittedly, you may see fewer in volume at first, but each lead will likely be “warmer.” The key is to make the most of it with an excellent user experience (easy contact forms, quick demos, etc.).

GEO also offers an opportunity to strengthen your brand image. By shaping your content to be as useful, clear, and reliable as possible, you improve the perception of your business across all audiences: AIs, human readers, and existing clients who appreciate your educational resources.

This pursuit of quality can push you to become a thought leader in your sector, producing resources that everyone cites (including competitors). Ultimately, it’s a win-win: more visibility in AI-generated answers, and more respect from your professional community.

Finally, GEO forces you to innovate and stay informed about search engine evolution. In doing so, you build an internal culture of marketing agility. Search engines will evolve, AIs too—but if your team is used to adapting (e.g., by testing new tools to track presence in AI outputs or experimenting with innovative content formats), you’ll always stay one step ahead.

Consider GEO as a learning ground that will make you more competitive overall—even beyond AI, across all aspects of digital marketing.

Conclusion: GEO and SEO hand in hand toward the future of search

GEO is quickly moving from being a buzzword to becoming an essential skill for digital marketers. Don’t think of it as a radical break from SEO, but rather as its natural extension in a new technological context. Sure, tomorrow’s search engines won’t just display a list of blue links—they’ll deliver an AI-written response. However, to build that response, AIs will still need high-quality content produced by humans—that is, by you, your company, your writers.

By following the recommendations shared in this article, you’ll start adapting your content so it can seamlessly travel across formats and interfaces. Whether someone searches on Google or asks an open question to ChatGPT, your goal is for your expertise to surface at the right moment. GEO and SEO don’t cancel each other out—they complement one another. By investing in both, you ensure maximum visibility.

Remember, we are still at the beginning of this transition. GEO best practices will continue to evolve as we better understand how AIs select and present information. Stay curious, keep learning—read studies, test queries yourself on different platforms, follow feedback from other specialists. This adaptability will be your strongest ally.

In the meantime, it’s not too early to take action: audit your existing content, identify quick GEO wins (a FAQ page to structure, an expert article to update, a guest post to publish on an influential site…), and get moving. Every small step toward clearer, more reliable, and better-structured content is a step toward greater visibility in the generative AI landscape.

SEO has long been king, but a new era is opening with GEO. By intelligently combining both, you’ll be ready to dominate across all engines—whether they run on links or artificial intelligence.

Don’t wait for competitors to get ahead: the time is now to optimize for the next generation of leads!

For more on best practices for integrating AI into your business, check out our complete guide.

GEO (Generative Engine Optimization): what is it?

/in /by

We increasingly hear people saying that “SEO is dead, long live GEO” – especially among those who enjoy proclaiming the dawn of a revolution. Should we panic and throw away years of best practices in search engine optimization? Certainly not. In reality, traditional SEO (Search Engine Optimization) isn’t “dead”: it’s evolving. A new discipline is emerging, complementary to SEO, designed for AI-powered search engines like ChatGPT or Perplexity. Its name: GEO, short for Generative Engine Optimization.

In this article, we will demystify the concept of GEO. You’ll learn what Generative Engine Optimization is, how it differs from traditional SEO, and why it is becoming essential with the rise of generative AI. The goal is to give you clear explanations (even if you’re not a technical expert), along with concrete examples.

Get ready: generative AI is changing the game of online search. With a little curiosity and the right practices, you can embrace this innovation confidently and turn it into a competitive advantage.

What does “GEO” mean ?

Generative Engine Optimization (GEO) refers to all content optimization techniques aimed at getting your brand or web pages to appear in the answers generated by AIs (also called conversational engines) such as ChatGPT (search mode), Google SGE (Search Generative Experience), Perplexity.ai, Bing Chat, Claude, or You.com. In other words, where traditional SEO seeks to rank you at the top of search results, GEO aims to integrate you directly into the response that an AI formulates for the user.

Put simply, the goal of GEO is no longer just to appear at the top of a list of links, but to be included, referenced, and cited within the answer an AI provides to the user. It’s no longer about earning a ranking position, but about earning a citation from the generative engine.

This shift in objective changes the very nature of optimization: you’re no longer creating content solely for a “static” search engine (which merely indexes and ranks pages), but for an AI model that understands, reinterprets, synthesizes, and redistributes information.

GEO vs SEO: what are the differences?

 

 

GEO and SEO share the same fundamental goal: making your content visible and relevant in order to answer user queries, regardless of the underlying technology. In fact, AI-driven engines ultimately pursue the same purpose as Google or Bing: providing a satisfying answer to the user’s question. To do so, they must index content, evaluate its quality and relevance, and present a meaningful result.

On these points, the foundations of SEO (quality content, relevance, user experience, and the E-E-A-T principles – Experience, Expertise, Authoritativeness, Trustworthiness) remain fully applicable to GEO. In other words, GEO and SEO “follow the same logic” of optimization, applied to new types of responses.

On the other hand, several key differences set GEO apart from traditional SEO:

  • Final objective: in SEO, the goal is to improve your site’s ranking on the search engine results page (SERP) of a traditional search engine. In GEO, the objective is to be cited and integrated into the AI-generated response. Your content therefore needs to be sufficiently informative and trustworthy for the AI to select it and include it in its synthesis.

 

  • Relevance criteria: traditional SEO is heavily focused on keywords (the right words in your pages to match typed queries) and on popularity signals such as backlinks. GEO, however, relies more on the semantic and contextual understanding of AIs. Since conversational queries are longer and more nuanced than a string of keywords, GEO-optimized content aims to cover a topic in depth and with context, rather than simply repeating an exact keyword. (Queries addressed to AIs already average 23 words, compared to 4 words for traditional searches — a sign that they are much more detailed and natural.)

 

  • Result format: a search engine like Google lists 10 blue links (plus snippets, images, etc.), while a generative engine provides a single written answer, often citing its sources through references embedded in the text. Ranking matters less than the relevance of the content itself: if your page is the best source, the AI may cite it as the first reference in its explanatory text, or even draw from it without always displaying a direct link. There is no visible #1 position in the same way — you are either included in the answer, or you remain invisible.

 

 

  • Authority and trust: in traditional SEO, a site’s popularity (domain authority, number and quality of backlinks) is a major ranking factor. In GEO, AIs will favor content that demonstrates authority through reliability and sources. For example, generative models give weight to pages that cite studies, data, or trusted sources (government sites, universities, recognized publications). Content that backs up its claims with verifiable facts will be considered more trustworthy by the AI. In this sense, including source citations or statistics in your content can significantly increase your chances of being picked up in a generative answer — up to +40% visibility according to a recent study.

 

  • Content optimization: SEO has trained you to polish your titles, meta descriptions, keyword density, etc. In GEO, the priority is to ensure that content is clear, educational, and well-structured. AIs love texts that are easy to break down: short paragraphs, bullet lists, FAQs, comparison tables… This type of structure allows the model to better understand and extract information without distorting it. Tests have shown that generative engines are more likely to reuse content structured as bullet points or step-by-step guides. In short, well-structured and segmented content has a higher chance of being “pulled in” and reused than unstructured text.

 

  • Performance measurement: in SEO, you look at metrics such as click-through rate (CTR) on the SERP, organic traffic, etc. In GEO, the focus shifts to new indicators like the reference rate: how often your brand or your content is cited as a source in AI-generated answers. This represents a new way of thinking about brand visibility. Already, tools are emerging to track these metrics: for example, Ahrefs and Semrush are developing features to monitor brand mentions in AI responses and analyze how your content is being reused.

In summary, GEO is not a “different SEO” opposed to the old one. It is a natural extension of optimization practices adapted to a new format of results. The two approaches are complementary: the best digital strategies will integrate both traditional SEO to stay visible on Google/Bing, and GEO to gain visibility on AI-driven platforms.

In fact, content that performs well in SEO often has a good chance of performing well in GEO too, as long as it is clear and up to date — and conversely, what benefits GEO (quality, clarity) often also strengthens traditional SEO. So there’s no need to panic: the goal is to adapt and enrich your practices, not to start over from scratch.

Why is GEO becoming essential in the era of ChatGPT and generative AI?

 

You might be thinking: “All of this is interesting, but how big is the phenomenon really? My clients are still using Google…” It’s true: for now, searches through generative engines still represent only a small share of total web searches. For example, a Similarweb study shared by Rand Fishkin estimated that ChatGPT (in Search mode) accounted for just 4.33% of queries compared to traditional search engines.

In other words, the majority of traffic is still on Google and the like at this point. But beware: although this share is modest, it is growing rapidly — and more importantly, the underlying trend is changing the way users find information.

 

 

Here’s why GEO deserves your attention right now:

  • The rise of conversational and voice search: AI assistants (from Siri to Alexa, and even chatbots in our browsers) are becoming increasingly capable of responding in natural language or even out loud. Younger generations sometimes prefer asking a bot directly rather than typing a traditional query. Generative engines provide more natural, contextual answers than the old assistants, which could further democratize voice search through smartphones and connected devices. If voice search takes off thanks to AI, you’ll want your business to be part of the spoken answers these tools deliver.

 

  • Younger users adopt new interfaces faster: there’s a generational difference in adoption. Younger internet users readily embrace tools like ChatGPT for their searches, while older generations remain more faithful to Google. If your target audience includes digital natives, it’s strategic not to miss the GEO opportunity. Being absent from AI responses could mean missing part of your future audience.

 

  • The convenience of direct answers: “Zero-click search” at its peak: let’s admit it — we all love getting an immediate answer to a simple question without digging through 10 websites. Generative engines excel here: for many basic or informational queries, they provide the answer directly, saving users from hopping across multiple pages. This fulfills a need that traditional SERPs only partially addressed. The consequence for businesses: more and more queries (especially informational ones, “top 10” lists, tutorials, etc.) may no longer generate clicks to websites, since the AI summarizes everything itself. Early signs already point to a decline in clicks on certain “top of funnel” content (general blog posts, lists, how-to articles) as AI answers replace them. Failing to adapt your strategy could mean losing visibility and organic traffic on those queries — to competitors the AI decides to highlight.

 

  • More qualified leads thanks to AI: while the volume of traffic from AI-generated answers is still smaller today, the quality of interactions and conversions can be higher. Experts note that users treat AI almost like a trusted advisor, which increases the conversion rate of AI-generated recommendations. In other words, if ChatGPT “recommends” your product or site, the user who clicks is already positively inclined — almost as if a friend suggested it. This can result in a stronger conversion rate (even if the total number of leads remains lower than through Google). That’s a promising sign: fewer clicks, but more qualified ones with higher potential to turn into customers.

 

  • An opportunity to get ahead of the competition: GEO is still new, and few businesses are actively optimizing for it. That means now is the time to be a pioneer. Showing up in AI answers before your competitors do allows you to gain valuable share of voice. As generative search grows in market share, you’ll already have a head start. Conversely, ignoring GEO now means risking playing catch-up later. Just look at how quickly giants like Apple are announcing the integration of AI engines (Perplexity, Claude) into products like Safari. The shift can accelerate overnight.

In short, GEO will take on an increasingly important role in online visibility strategies. While Google and traditional SEO remain indispensable today, it is wiser to start considering GEO right now as an essential strategic complement. As one expert summed it up: “GEO isn’t going anywhere, so keeping up and earning visibility in GEO makes sense. Just remember to manage GEO with SEO.”

In other words, don’t put all your eggs in the AI basket — but add this new lever to your marketing arsenal to ensure your visibility no matter what.

Challenges for lead generation and business visibility

From a company’s perspective, what does the rise of generative search actually change for lead generation and visibility? Here are the main challenges to keep in mind:

1. The potential decline of “traditional” organic traffic: as mentioned, if AI provides a complete answer, users click less on results. Fewer clicks to your site means fewer opportunities to capture leads via your forms, newsletters, etc. For example, a website relying on “How to do X” tutorials could see its traffic drop if users get the step-by-step directly from a chatbot. It therefore becomes crucial to anticipate this decline by investing in the channels where users now find their answers (AI platforms) to capture that contact in a different way.

2. The importance of being in the answer (otherwise, you don’t exist): on Google, you could still hope for traffic being the #3 or #4 result, or even on page 2 for niche queries. With a generative engine, either your brand is cited in the answer, or it’s invisible. It’s almost all or nothing. A concrete example: imagine a user asks ChatGPT, “What is the best project management software for a small business?” The AI will likely list 2 or 3 recommended tools, each with a short explanation. If your software is on that list, you reach the prospect; if not, the prospect won’t even know you exist. There’s no page 2 where they might discover you. For lead generation, being included in the AI’s answer is therefore critical, especially for “product discovery” queries.

3. A shorter user journey: AI can act as a filter by presenting only a few options instead of dozens of results. This means the discovery journey is shorter and concentrated on the first suggestions. If your offer is among them, the potential lead arrives already further along in their decision-making (the AI has done the initial comparison work for them). They are warmer, more convinced. But if you are absent from the AI answer, they won’t come at all. GEO thus impacts the conversion funnel: it shortens the information-gathering phase and can deliver more qualified prospects directly at the consideration/purchase stage.

4. Rethinking conversion content: less traffic doesn’t necessarily mean fewer conversions if the remaining traffic is of higher quality. You’ll likely need to optimize your landing pages and conversion content (product demos, free trials, etc.) to welcome these AI-driven visitors. They often arrive through a direct link in the answer (e.g., “Learn more on [your brand]’s website”). They come looking for more detail or proof after the AI’s recommendation. Your site must therefore be flawless to quickly convince them: clear content, social proof, visible calls-to-action. GEO may bring you fewer visitors, but you need to convert them better.

5. Measuring and monitoring your presence in AI: as noted, new metrics and tools come into play. You’ll need to monitor whether and how your brand is mentioned by AIs. Some companies are already conducting qualitative monitoring: regularly asking sample questions to ChatGPT, Bing Chat, Perplexity, etc., to see if their content or products are cited. Specialized tools are emerging (e.g., AI search graders) to make this easier. This monitoring is critical: it helps you understand how AI portrays your company (accuracy of information, positive/negative sentiment, etc.) and adjust your content strategy accordingly. It’s somewhat the equivalent of SEO ranking tracking, but transposed into AI-generated answers.

In summary, GEO presents a dual challenge for businesses: not losing the visibility already gained through traditional SEO, by continuing to feed that channel, while also building visibility on the new generative channels. Those who succeed in this transition can turn the threat – less “easy” organic traffic – into an opportunity : more qualified leads, and a competitive edge in a still underexploited space.

For more general insights on integrating AI into your business, check out this article.

NotebookLM and its alternatives: information retrieval at the heart of business needs

/in /by

NotebookLM, Microsoft AI Search, RAG… these are names I keep hearing in companies. Indeed, I’ve noticed a recurring challenge: employees struggle to quickly find the information they need. Whether it is an HR chatbot to query internal policy, a tool to retrieve business procedures, or an assistant to navigate massive document databases, the demand is always the same: how to access the right information, without wasting time?

This need, though simple in appearance, faces several difficulties: heterogeneity of data (PDFs, emails, internal files), multiplicity of sources (Drive, SharePoint, intranet), and lack of user-friendly tools for non-technicians.

The different possible approaches


Faced with this challenge, several technological strategies are emerging. In this article, we will focus on three major solutions that come up most often in business projects: RAG (custom approach), Microsoft AI Search (in the Microsoft ecosystem), and NotebookLM (in the Google ecosystem).

 

1. RAG (Retrieval-Augmented Generation)


RAG (Retrieval-Augmented Generation) is today a widely used approach. It consists of enriching a language model with a document base: the system first looks for the relevant passages (retrieval), then generates a contextualized answer from these excerpts (generation). It is an architecture that can adapt to many use cases: internal chatbots, advanced search engines, business assistants, etc.

 

  • Advantages: great flexibility, independence with respect to a particular ecosystem (Google, Microsoft…), advanced customization according to the needs of the company, and total control over how the data is processed and delivered
  • Disadvantages: more technical implementation, requiring skills in AI and engineering (building pipelines, hosting, managing embeddings and infrastructure). This also implies non-negligible maintenance and evolution costs.

 

2. Microsoft solutions : Azure AI Search 


For companies already integrated into the Microsoft ecosystem (SharePoint, Teams, Azure), Microsoft AI Search (formerly Azure Cognitive Search) is a natural option. This tool is a managed AI-based search service, which allows indexing varied data (Office documents, PDFs, SQL databases, images…) and making them accessible via full-text search, semantic vector search, or even AI enrichment pipelines (OCR, entity extraction, translation, etc.).

 

  • Advantages: smooth integration with the Microsoft ecosystem, high security and compliance, large processing capacity (scalability), and compatibility with RAG to build conversational assistants.
  • Disadvantages: strong dependence on Microsoft, more technical configuration than turnkey solutions, less flexibility than custom development, and costs that can quickly rise with advanced options.

 

3. Google’s NotebookLM 


NotebookLM is Google’s tool designed to facilitate research and synthesis of information in your own documents. Unlike a generalist chatbot, it specializes in the documentation you provide it (Google Docs, PDFs, notes, etc.), which makes it a true personalized research assistant.

It stands out for its ease of use: no need for complex pipelines or technical configuration, you just import your documents into a “notebook” and the AI is immediately able to answer your questions, summarize content, or generate structured syntheses. NotebookLM is also very easy to use via a personal Gmail address (see tutorial below).

 

Advantages:

  • Intuitive interface, designed for non-technicians.
  • Direct integration with Google Drive and Docs.
  • Summarization and contextualization capabilities very useful to synthesize reports, meeting notes, or internal policies.
  • Accessible to anyone with a Gmail address (ideal for quick testing).

Disadvantages:

  • Limited to the Google ecosystem: less suitable if your documentation is stored elsewhere (SharePoint, internal databases…).
  • Fewer customization possibilities than a custom RAG approach.

Tutorial to quickly and easily test NotebookLM

  1. Log in with a Google account
  • Go to notebooklm.google and log in with your Gmail address. The tool is available for free in trial version.

    2. Create a new “notebook”

  • Click on “Create new notebook” to start.
  • Give a title to your notebook (e.g., “Employment leave: rights and obligations”).

   3. Add sources

  • Import documents (e.g., public documents available on the UK government website GOV.UK about statutory leave and absences).
  • You can notably upload PDF files or copy-paste the content of an official page into a Google Doc, then add it to your notebook.

   4. Ask your first questions

In the chat bar, test queries such as:

  • “What is the statutory minimum holiday entitlement in the UK?”
  • “What types of leave are available for family events, such as a wedding or a bereavement?”
  • “What are the employee rights in case of statutory sick leave?”

NotebookLM will display the answer, indicating the exact passages of the document it relies on.

 

 

Explore the features

  • Automatic text summary: generate a clear digest of all rules related to leave and absences.
  • Video summary: generate an explanatory video of your documents’ content.
  • Mind map: NotebookLM can organize types of leave in a visual map: paid, family, parental, sick, etc.

  • Automatically generated podcast: NotebookLM can transform the content into an audio discussion between two synthetic voices. This is an original and playful feature, but more anecdotal than truly useful in a professional setting.

👉 In a few minutes, even a non-technical user can transform sometimes dense legislative texts into clear answers, usable summaries, or visual maps easy to understand.

Which choice for which company?

  • You are 100% Google Workspace: NotebookLM is a fast, effective, and frictionless solution.
  • You are anchored in Microsoft 365: Cognitive Search / AI Search is the natural extension.
  • You are looking for a fully custom solution: RAG, with its personalized pipelines, will be the best option, at the cost of greater complexity.

Conclusion


Information retrieval has become a major challenge in a context where employees are overwhelmed with documents and data. Solutions exist to address it, whether turnkey (NotebookLM, AI Search) or more personalized (RAG).

The choice is not just about opting for the most sophisticated tool, nor limiting oneself to the one that integrates most easily. The best option is often found in the balance between performance and integration: the tool must fit into the company’s environment and culture while generating real functional gains.

 

For further insights on integrating AI into your organization, explore our comprehensive guide.

Introducing COPILOT(): Excel’s useful AI-powered function !

/in /by

What if writing complex formulas in Excel became as simple as writing a sentence? Introduced in August 2025, the new COPILOT() function transforms the traditional use of spreadsheets by directly integrating AI — and without scripts.

 

1. What is the COPILOT() function?

It is a native Excel formula — just like SUM or IF — but this time powered by artificial intelligence. In practice, you enter a natural language prompt, optionally adding cell references, and Excel generates an intelligent and appropriate response.

The COPILOT() function is integrated into the calculation engine: results update automatically when data changes.

 

 

2. What is it for? Some practical use cases

 

Brainstorming:

The function can be used to support brainstorming directly within an Excel sheet. It generates ideas for marketing campaigns, new product features, or SEO keywords from a simple description. It can also rephrase a message to make it clearer or adapt its tone as needed.

Summary generation:

COPILOT() condenses large datasets or long texts into clear and concise summaries. It highlights trends, simplifies complex calculations, or produces understandable explanations, ideal for preparing reports or presentations.

 

 

 

Data classification:

No more exports to other tools: the function directly classifies customer feedback, support tickets, or survey responses within Excel. It can perform tagging or sentiment analysis to better organize and exploit the information.

 

 

List or table creation:

The function automatically generates lists and tables tailored to your model. This can be useful to quickly build a test dataset, list industry examples, or sketch out a project plan. The results spill into the Excel grid, in ready-to-use columns and rows.

 

 

COPILOT() can be nested within other Excel functions such as IF, SWITCH, LAMBDA, WRAPROWS… to automate more advanced dynamic processing.

3. Availability and access

The COPILOT() function is currently available to Microsoft 365 Copilot subscribers, in the Beta channel versions:

  • Windows (from build 19212.20000 / v. 2509+)
  • Mac (build 25081334 / v. 16.101+)

A Web version is planned through Excel for Web’s “Frontier” program (learn more).

4. Limitations and recommendations

Microsoft specifies that the COPILOT() function relies on AI and may provide incorrect answers. To ensure reliability and responsible use, it is recommended to avoid using it in the following cases:

  • Numerical calculations: instead, use Excel’s native formulas (SUM, AVERAGE, IF…)
  • Answers requiring external context: the function only has access to the prompt and referenced cell ranges
  • Data lookups: prefer XLOOKUP for reliable searches.
  • High-stakes scenarios: financial reports, legal or regulatory documents.
  • Recent or real-time data: the function is non-deterministic and may return different results upon recalculation. Moreover, the model is limited to knowledge prior to June 2024.

 

The COPILOT() function marks a new step in spreadsheet efficiency and simplicity. It does not replace Excel’s classic formulas, but complements them by opening new possibilities through natural language. Its main value lies in idea generation, data summarization, or text content classification — tasks often tedious to perform manually. Like any AI-based technology, it has limitations and requires thoughtful use, particularly for accuracy or sensitive context. But when wisely integrated into workflows, it can make Excel more accessible, more creative, and above all more productive on a daily basis.

 

For more information about Copilot & Excel, see this article.

For more general information about integrating AI into your company, see this article.

The “Analyst” agent: where Excel meets avanced data analysis

/in /by

1. What is the “Analyst” agent?

Analyst is an intelligent conversational agent integrated into Microsoft 365 Copilot, designed to assist you in data analysis without requiring complex technical skills. It acts like a “qualified data analyst” at your fingertips, capable of processing and synthesizing your information seamlessly.

Built on an advanced reasoning model, post-trained on GPT o3-mini, this agent simulates human reasoning strategically: it formulates hypotheses, tests, refines its results, and corrects its errors, much like a real-time data scientist.

 

2. What is the purpose of the “Analyst” agent?

The Analyst agent was designed to simplify and accelerate analysis tasks, reducing the technical barrier often associated with data manipulation. Where previously one had to juggle complex formulas, macros, or even scripts, the agent now enables results in natural language, with a much more intuitive approach.

Here are some of its main uses:

  • Consolidate and analyze scattered data
    It can instantly aggregate information from multiple Excel files, databases, or CSVs. It then becomes a central hub for a unified view, without tedious manipulations or multiple exports.

  • Transform raw data into readable reports
    Whether it’s thousands of rows of sales or customer feedback, it can turn them into trends, key indicators, or understandable visualizations. It doesn’t stop at numbers: it can also generate narrative summaries to accompany charts and contextualize results.

  • Produce scenarios and simulations
    It can create projections, such as estimating revenue growth or anticipating product demand. Thanks to its reasoning logic, it not only provides a raw number but also explains the assumptions made, helping to assess the relevance of the result.

  • A few concrete use cases

    • For marketing: analyze customer feedback, identify market segments, or detect emerging trends.

    • For finance: create simplified dashboards, calculate performance indicators, or simulate the impact of different budget strategies.

    • For operations: optimize inventory management, anticipate demand peaks, or track logistics KPIs.

In short, Analyst acts as an integrated data analyst assistant within Microsoft 365: it reduces the time spent preparing and cleaning data, and increases the time devoted to interpretation and decision-making.

Example:

 

3. Availability and integration in Microsoft 365 Copilot

Analyst is part of the reasoning agents — alongside the “Researcher” agent — launched for the first time under the Frontier program, in early access starting April 2025 (learn more).

They are now generally available to all Microsoft 365 Copilot subscribers. You will find it directly in Copilot Chat, ready to use across all supported platforms (web, desktop, mobile).

It is part of a broader trend: the arrival of AI agents specialized in automating complex tasks in Microsoft 365 (learn more).

4. Why is this important?

  • Increased productivity: Analyst transforms hours of manual data manipulation into minutes of automated analysis, even for advanced needs.
  • Enhanced accessibility: no need to be a data scientist or Excel expert to get powerful insights. The agent makes data analysis more fluid, intuitive, and collaborative.
  • Security and consistency: connected to Microsoft Graph, it operates using your work data — emails, files, chats… — while complying with internal security and privacy standards.

5. A word on advanced reasoning

The major strength of Analyst lies in its step-by-step reasoning approach (chain-of-thought). Unlike a classic assistant that merely generates an immediate answer to a question, this agent adopts an approach closer to that of a real data analyst.

Concretely, it:

  • breaks down the problem into sub-questions,
  • formulates hypotheses,
  • tests different scenarios on the data,
  • adjusts its calculations based on results,
  • and explains its reasoning transparently.

This method makes it particularly suitable for contexts where data is complex, incomplete, or scattered.

Conclusion

Analyst marks an important step in the evolution of productivity tools. Where Excel, Power BI, or other solutions previously required solid technical expertise, the agent now provides access to powerful analysis in natural language. Its strength lies in its ability to reason, handle heterogeneous data, and deliver actionable insights as trends, visualizations, or synthetic reports.

It is not about replacing human analysts or Excel’s advanced features, but rather making data exploration and interpretation more accessible to a wider audience. As a complement to experts, Analyst can speed up the preparation of analyses, highlight interpretation paths, and free up time for higher-value tasks.

For more information on Copilot & Excel, see this article.

For more general information on AI integration in your company, see this article.