Why Act Now?
Generative AIs like ChatGPT, Perplexity, and Gemini are no longer technological curiosities — they've become full-fledged search interfaces. ChatGPT now exceeds 800 million weekly active users, and Gartner estimates that traditional search engine volume will drop by 25% by the end of 2026, in favor of conversational assistants.
For content creators, this poses a concrete problem: generative engines don't list links — they synthesize and rephrase. Your content doesn't just need to rank well, it needs to be understood, cited, and rephrased by an AI.
That's where GEO (Generative Engine Optimization) comes in. The term was formalized in November 2023 by researchers from Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi in a study published on arXiv, then presented at the KDD 2024 conference. Their main finding: the right GEO techniques can improve a website's visibility in generative responses by up to 40%. Even better, lesser-known sites gained up to 115% more visibility by adopting these methods — meaning GEO is also an opportunity for players who don't yet dominate traditional SEO.
Here are 5 concrete actions you can implement today.
1. Structure Your Data with JSON-LD
Structured data is one of the few technical signals you directly control that generative AIs can readily exploit. It allows models to extract precise information about your business, content, and products without having to infer context from free text.
At a minimum, implement the following schemas, following the official Schema.org documentation:
- Organization: name, logo, address, social media profiles
- Article: title, author, publication and update dates
- FAQPage: your most frequently asked questions and answers
- Product: for your product pages, including price and availability
AIs use these tags to build reliable answers. A well-filled FAQPage schema, for example, significantly increases the chances that your Q&A will be directly incorporated into a ChatGPT or Perplexity response. Google provides a structured data validator to check your implementation before publishing.
Note: JSON-LD must be consistent with the visible content on the page. A discrepancy between the two is a negative signal for both traditional search engines and AIs.
2. Adopt a Factual, Source-Backed Tone
This is one of the strongest findings from the Princeton study: Statistics Addition (adding statistics) and Quotation Addition (adding sourced citations) are the two techniques that most improve visibility in generative engines — far outperforming keyword stuffing, which actually proves counterproductive.
Why? Because LLMs are trained to favor sources that demonstrate authority. A quantified, sourced claim is easier to verify, extract, and cite than a vague assertion.
Before: "Our tool is the best on the market."
After: "Our tool analyzes 15 distinct GEO criteria. Across a panel of 500 supported websites, we observe a median AI visibility improvement of 40% within 90 days, consistent with findings from academic research (Aggarwal et al., 2024)."
The difference isn't just stylistic — it's a difference in trust signal for the model.
A few practical rules:
- Cite your sources with a hyperlink to the original source (study, institutional report, mainstream news article)
- Prefer precise figures over approximations ("43%" rather than "about half")
- Mention the author or source organization directly in the text, not just in footnotes
3. Create "Direct Answer" Content
Generative engines operate in question-and-answer mode. They search your content for passages they can extract and rephrase to answer a specific query. If your content isn't structured for that, it will be overlooked in favor of a better-organized competitor.
In practice, this means:
Question-format headings: an H2 like "What is GEO?" is more usable for an AI than a vague title like "Introduction." Think about what a user might type into ChatGPT or Perplexity.
Short, direct introductory paragraphs: ideally, the first sentence after a heading should already provide the essential answer. Details and nuances come next. This format, sometimes called BLUF (Bottom Line Up Front), is naturally favorable for automatic extraction.
Bullet lists for steps and criteria: they are more easily parsed than enumerative lists written in prose.
An article summary at the top (TL;DR): more and more websites add an "In summary" or "Key takeaways" block at the beginning of the page. This is a strong signal for AIs, which often look for a synthetic anchor before diving into the details.
The goal isn't to write for robots at the expense of humans — it's to write for humans clearly enough that robots can follow along.
4. Regularly Update Your Content
Freshness is an active selection criterion for AI models. Content dated from 2021 with no visible updates will be systematically deprioritized compared to similar content updated in 2026, all else being equal. This is particularly critical in fast-evolving fields like AI, tech, or regulation.
This isn't just intuition: according to Gartner, 25% of traditional queries will migrate to AI interfaces by 2026 — and these interfaces have very low tolerance for outdated information, because they stake their credibility on the answers they generate.
Here's an effective update routine:
- Display a readable last updated date, separate from the initial publication date
- Audit your high-traffic articles every quarter: statistics, cited tools, and links should remain valid
- Update key figures as soon as a new authoritative source becomes available
- Use IndexNow to instantly notify Bing, Yandex, and compatible engines of your changes — this open-source protocol is free and reduces the time before re-indexing
An update shouldn't be cosmetic. Changing two words just to update the date has no effect. What matters is adding substantive value: new data, new sections, correcting information that's become inaccurate.
5. Optimize Your Topical Internal Linking
Generative AIs assess a site's topical authority holistically, not on a page-by-page basis. A site that covers a topic in depth — with interconnected articles, complementary angles, and logical progression — will be perceived as a reference source. A site with a single good isolated article will not.
This principle is known in SEO as topical authority, and it applies even more strongly to GEO. Language models are trained on corpora where certain domains are represented by many sources that are consistent with each other. Your site should aspire to be one of those consistent sources on your topic.
In practice:
Create content clusters: group your articles around master themes. If you publish about GEO, you should cover structured data, factual content, domain authority, freshness, measurement tools — and link all these articles together.
Develop comprehensive pillar pages: a pillar page is an exhaustive article on a given subject that serves as a hub to more specialized articles. It's longer than average, covers the topic from every angle, and is regularly updated.
Link your articles contextually: an internal link is more useful for an AI (and for a human reader) when anchored on precise, descriptive text, not on "click here." Prefer anchors like "our guide to implementing JSON-LD" over a bare link.
Seek external mentions: Reddit, LinkedIn, and YouTube are among the most frequently cited sources by LLMs in their responses. Distributing your content on these platforms, answering questions in your domain, and creating authentic discussions — all of this strengthens your presence in the corpora that AIs rely on.
Bonus: Measure Your Progress
Optimization without measurement is flying blind. Several GEO-specific metrics are emerging today: citation rate in AI responses, share of voice on key queries, and conversion rate of traffic from AI platforms — the latter being estimated at 4.4x higher than traditional SEO.
Use vIAsibility to regularly audit your GEO score, track your progress, and identify priority areas for improvement. Tools like Semrush AIO and Profound also let you monitor your brand's visibility on ChatGPT, Perplexity, and Google AI Mode.
Conclusion
GEO isn't a revolution that erases everything you know about SEO — it's a logical extension of the same principles: create useful, well-structured, credible, and up-to-date content. What changes is the final reader: it's no longer just a ranking algorithm — it's a language model that will synthesize your content and integrate it into a response to a user.
The 5 levers presented here — structured data, factual sourced tone, direct answer format, content freshness, and topical authority — are validated by both academic research and real-world feedback. They are achievable without a major technical overhaul.
Start with the two that will have the most impact on your specific site, measure the results over 90 days, and iterate. The best GEO strategy is one that adapts to your audience and your domain — not a copy-pasted template.