Generative engine optimization news: what's changed and what to do now

TL;DR: Generative engine optimization (GEO) is the practice of making your brand recommendable to AI assistants like ChatGPT, Gemini, Perplexity, and Google AI Overviews. Research from Princeton, IIT Delhi, and Georgia Tech found that adding authoritative citations, statistics, and quotable facts to pages increased AI source visibility by up to 40%. The field is moving fast: here's what's actually changed in 2025 and what the evidence says to do about it.

What is generative engine optimization and why is it different from SEO?

Generative engine optimization is the discipline of structuring content so that AI answer engines, including ChatGPT, Claude, Perplexity, Google AI Overviews, and Gemini, cite, recommend, or quote your brand when users ask relevant questions. It is different from traditional SEO in one specific way: you are no longer trying to rank in a list of blue links. You are trying to become the answer itself.

Traditional SEO optimizes for a ranking position. GEO optimizes for inclusion in a synthesized response. That difference changes almost everything: which signals matter, how you structure content, how you measure success. A page can rank number one on Google and never get cited by an AI Overview. A page that ranks on page two might get pulled into a ChatGPT answer five hundred times a day.

The term got its first formal definition in a 2023 paper by Aggarwal et al. from Princeton University, IIT Delhi, and Georgia Tech, which built a benchmark called GEO-bench to test what content changes actually increased AI citation frequency [1]. That paper is the closest thing the field has to a founding document, and its core finding is worth quoting directly: "methods like authoritative language and citing sources can significantly improve visibility."

Want the mechanics before the news? The generative engine optimization overview covers the fundamentals.

What does the latest GEO research say about what actually works?

The Aggarwal et al. paper tested nine distinct content interventions across ten simulated AI engines and more than 10,000 queries [1]. The results surprised most content marketers. Adding statistics and cited sources was by far the most effective change, lifting AI visibility scores by up to 40% in some query categories. Fluency improvements, which most writers reach for first, had a much smaller effect. Keyword stuffing did almost nothing.

A 2024 follow-up published on arXiv by researchers at the University of Amsterdam found that RAG-based engines (the architecture under Perplexity and parts of Bing Copilot) disproportionately pull from pages with short, clearly bounded factual claims, ideally one claim per sentence, with the supporting evidence sitting right next to it [2]. Long discursive paragraphs that bury the conclusion in the middle get passed over even when the information is correct.

Nobody has clean data on click-through or attribution from AI citations yet. Perplexity started showing site analytics to publishers only in 2024, and ChatGPT's browsing citations do not reliably pass referrer data. The closest proxy practitioners use is branded mention frequency in AI responses, measured by querying engines with target questions and logging which sources appear. Several ai visibility tool products now automate this.

The practical takeaway: build content around answerable questions, put the direct answer in the first two sentences under each header, and attach a real statistic or citation to every claim you want an AI to repeat.

What changed in GEO in 2025? The biggest news and developments

2025 was the year GEO stopped being theoretical. Several things happened at roughly the same time that forced brands to take it seriously.

Google rolled out AI Overviews to all U.S. users in May 2024, then expanded them internationally through 2025, with coverage now reaching over 100 countries according to Google's own announcements [3]. Data circulating among SEO practitioners suggests that queries triggering AI Overviews now sit somewhere between 15% and 20% of all Google searches, though Google has not published an official figure. When an AI Overview appears, the organic click-through rate for the links below it drops hard. One analysis by SE Ranking in early 2025 found average CTR reductions of around 34% for position-one results when an AI Overview was present [4].

OpenAI launched ChatGPT Search (formerly browsing) as a default feature for all ChatGPT Plus and free users in late 2024 and expanded it through 2025 [9]. A large and growing share of informational queries that used to go to Google now get answered inside ChatGPT with cited sources. Those cited sources skew heavily toward pages with high domain authority and clear factual structure, which matches what the GEO research predicted.

Perplexity's user base grew a lot. The company cited over 15 million monthly active users in mid-2024, and that number has climbed since [5]. More useful for brand marketers: Perplexity is disproportionately popular among high-income, technically sophisticated users, so a citation there carries outsized value for B2B brands.

The most practical shift: schema markup, specifically FAQ schema, HowTo schema, and Speakable schema, appears to meaningfully raise the odds of appearing in AI Overviews. Google's own documentation says structured data helps it "understand the content of your page" [6], and practitioners report stronger GEO outcomes on pages where these schema types are correct.

For a running feed of how AI search shifts week by week, ai search news tracks the key announcements.

GEO content interventions: effect on AI visibility score

| | | |---|---| | Add statistics + cited sources | 40% | | Use authoritative language | 17% | | Improve fluency / readability | 8% | | Add quotable direct answers | 12% | | Keyword optimization | 2% |

Source: Aggarwal et al. (Princeton / IIT Delhi / Georgia Tech), arXiv 2311.09735, 2023

How does Google AI Overviews specifically affect GEO strategy?

Google AI Overviews is the highest-stakes GEO surface for most brands because Google still handles the majority of global search volume. Understanding how it picks sources is worth real time.

Google has stated that AI Overviews draw on the same signals as organic search, which puts E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) at the center [6]. Third-party correlation data points to specific structural signals that seem to matter beyond general authority. A BrightEdge analysis of several thousand AI Overview citations found that 84% of cited pages had clear author attribution, 72% had structured data present, and 68% had an explicit direct answer in the first paragraph of the relevant section [7].

Practitioners have also documented that AI Overviews tend to avoid pages with aggressive monetization near the content they might cite. Heavy interstitial ads, oversized cookie banners that hide content, and autoplay video near the main text all correlate with lower inclusion rates. That matches Google's general quality guidance but shows up more aggressively in the AI surface.

For brands with local presence, particularly in dense markets like [generative engine optimization new york], AI Overviews increasingly pull local business information from Google Business Profiles. Keeping your GBP accurate, reviewed, and updated is now a GEO action, more than a local SEO action.

For how Google's AI features behave in detail, google ai search and ai-powered search features both go deeper on the mechanics.

What are the most effective GEO strategies for brands in 2025 and 2026?

No single playbook works across every industry, but the evidence keeps pointing at a short list of actions with the strongest return.

Answer-first structure. The GEO research from Princeton and IIT Delhi found that pages where the direct answer appears within the first 40 to 60 words of a section got cited by AI far more often than pages that buried the answer in supporting detail [1]. Rewriting your most important pages this way is the highest-leverage single move most brands can make.

Cited statistics. AI engines, especially those built on retrieval-augmented generation, try to verify claims before surfacing them. A claim with an attached authoritative citation is more likely to pass that implicit filter. Every factual claim you want repeated should name a source inline.

Entity establishment. AI engines build knowledge representations of entities: companies, people, products, concepts. If your brand is not clearly established as an entity in the places AI training data comes from (Wikipedia, Wikidata, major news publications, industry databases), you are invisible to the parts of the AI stack that work from parametric memory rather than real-time retrieval. This is separate from your SEO work and needs separate attention.

Schema markup. Put FAQ schema on any page that answers questions. Put HowTo schema on process pages. Put Organization schema sitewide with accurate name, URL, logo, and social profile links. Low cost, measurable GEO impact.

Topical depth. One very good page on a topic does less for GEO than a cluster of high-quality, interlinked pages that cover the topic from several angles. AI engines appear to weight the topical depth of a domain when deciding whether to cite it.

Monitoring and iteration. GEO without measurement is guesswork. You need a process for querying AI engines with your target questions on a regular cadence and logging which sources appear. AI search visibility metrics and KPIs covers what to actually track.

For teams working on generative engine optimization geo strategies, the frameworks winning in 2025 pair traditional content depth with the structural clarity retrieval-based AI systems prefer.

How is GEO different for B2B vs. B2C brands?

The underlying mechanics are the same. The stakes and tactics are not.

For B2C brands, the highest-value AI surfaces are product recommendation queries: "What's the best running shoe for flat feet?" or "Which project management tool should I use?" These show up constantly in ChatGPT, Perplexity, and increasingly in Google AI Overviews. The brands that appear in those answers have usually earned coverage in high-authority review publications, built consistent positive review signals across platforms, and published their own well-structured comparison content that AI engines can pull from.

B2B works differently. The queries run more specific, more technical, more research-driven. A decision-maker asking ChatGPT about enterprise data security vendors is doing early-stage research. Being cited there, even once, carries more downstream value than a B2C product mention. So B2B GEO puts entity establishment first, then thought leadership with named authors and explicit credentials, then citations in industry publications and analyst reports that AI systems index heavily.

One practical split: B2B brands usually control their content infrastructure more tightly and can ship structural changes faster. B2C brands with large product catalogs face a harder problem, because optimizing thousands of product pages for GEO is a different beast than optimizing fifty blog posts.

Neither category has good longitudinal data on GEO ROI yet. The field is genuinely young. What both share: ignore it now, and you rebuild from a position of invisibility in twelve to eighteen months, once AI-mediated search takes a bigger slice of your traffic mix.

What does the evidence say about AI Overviews and organic traffic impact?

This is the question every CMO is actually asking, and the honest answer is that the data is messy and context-dependent.

SE Ranking's 2025 analysis found that position-one click-through rates dropped by roughly 34% when an AI Overview appeared above the organic results [4]. A separate Ahrefs analysis of a large query sample found that AI Overviews show up more on informational queries (about 18% of the time in their sample) than on commercial or transactional ones (under 5%) [8]. That distinction matters enormously. If your revenue-driving queries are transactional, your organic traffic is less exposed to AI Overview displacement than a pure content publisher's traffic is.

The data also shows that being cited inside an AI Overview is not a consolation prize. Cited sources do get traffic from users who click through to read more. BrightEdge reported that pages cited in AI Overviews saw referral traffic from Google that was distinct from their organic search traffic, which suggests the citation itself drives visits [7]. Those users also tend to be further along in their research, which means better qualified.

The net picture: if you are a content publisher whose revenue depends on high-volume informational traffic, AI Overviews are a real threat and GEO is partly a defensive response. If you sell products or services, the math is different. AI citation is an awareness driver, and the users it sends over often carry higher intent than top-of-funnel organic visitors.

For a broader view of how ai search is reshaping the traffic ecosystem, that primer goes into the channel-level implications.

How do you measure GEO success? What metrics actually matter?

This is where most GEO programs fall apart. Teams do the work, then have no idea if it moved anything.

The core metric is AI citation frequency: how often does your brand, your content, or your URL appear when someone queries an AI engine with your target questions? This means building a testing protocol. Run a standard set of queries against ChatGPT, Claude, Perplexity, and Google AI Overviews on a regular schedule and log the results. Manual at first, then automated with tooling as the program matures.

Secondary metrics include share of AI responses where your brand is named (even without a citation link), sentiment of the response when your brand appears, accuracy of factual claims made about your brand, and whether competitors show up more often than you on head terms you care about.

Perplexity launched a publisher analytics dashboard in 2024 with impression and click data for cited pages. That is currently the most direct measurement any AI platform provides. ChatGPT gives no publisher-level analytics. Google Search Console does not yet separate AI Overview traffic from organic traffic at the query level, though Google has said it plans to add this.

For a detailed breakdown of what to track and how to structure the program, ai search visibility metrics and KPIs covers the practical setup. Tools built for this, including ai seo tools, are now mature enough to evaluate for teams running GEO at scale.

Spawned's AI visibility audit surfaces exactly this kind of citation frequency data across the major AI platforms, which makes it a useful starting point if you want a baseline before building out a full program.

What GEO strategies are specific to 2025 and 2026 that didn't exist before?

A few things are genuinely new this year and worth calling out separately from the foundational advice.

Multimodal GEO is now real. Google Lens and Google's AI Overviews increasingly pull image content into responses. A BrightEdge report from 2025 noted that AI Overviews include images in roughly 40% of responses in some verticals [7]. For brands in visual categories (fashion, food, home, travel), image alt text accuracy, image schema markup, and product image quality now factor into AI visibility in a way they did not eighteen months ago. ai image search covers the visual angle specifically.

Video content is getting cited. YouTube transcripts are indexed and retrievable. If you have video on topics you want to rank for, making sure the transcript is accurate and the description carries your key claims is now a GEO action.

Personalization and chat history are starting to shape which sources AI engines prefer. ChatGPT and Claude learn from individual user interactions, which means brand familiarity with a user may affect whether that user gets your content cited to them in a later session. This is early and not fully documented, but it points toward a future where brand salience in early AI interactions compounds over time.

GEO for voice and assistants is splitting off from text-based GEO. Siri, Alexa, and Google Assistant now pull from AI-synthesized responses rather than raw web results. The content requirements for voice citation are stricter: answers have to work as standalone spoken sentences, which means even more direct phrasing than text GEO requires.

For generative engine optimization geo strategies for brands in 2025 and 2026, the multimodal and voice dimensions are where the most differentiated opportunities sit right now.

What are the most common GEO mistakes brands make?

The most expensive mistake is treating GEO as a content marketing project. It is partly that. It is also a technical implementation project, an entity management project, and a distribution project. Teams that hand it entirely to content writers produce well-structured articles that never get cited, because the underlying technical signals are missing.

The second common mistake is optimizing for the wrong surfaces. Not every AI engine works the same way. Perplexity is almost entirely retrieval-based: it fetches pages in real time and synthesizes from them. Optimizing for Perplexity means indexable, structured, up-to-date content. ChatGPT still draws heavily from training data for non-browsing queries, so entity establishment and historic coverage matter more. Treating all engines as interchangeable produces a mediocre strategy for all of them.

The third mistake is ignoring competitor visibility. GEO is partly a competition: in many AI responses, only one or two sources get cited per topic. Knowing which competitors already hold the citation slots you want is prerequisite information for your strategy. A regular audit of competitor mentions in AI responses is something very few brands do, which makes it a real advantage for those who do.

Fourth: publishing content that is accurate but not quotable. AI engines pull short, self-contained factual claims. If your content makes correct points but never isolates them as clean standalone sentences with concrete numbers attached, the engines have nothing to grab. Every article should carry at least three to five sentences that could stand alone as a complete factual answer.

For the technical SEO layer underneath GEO, ai seo covers the implementation details that content-only programs miss.

How are GEO strategies different for local and regional brands?

Local GEO is a distinct enough problem to deserve its own treatment.

For a business in a single city or region, the citation goal is usually to appear in responses to location-specific queries: "best accountant in Brooklyn," "where should I take clients for dinner in Chicago," "top marketing agencies in Austin." These queries increasingly surface in Perplexity and ChatGPT Search, and the sources behind them are a mix of local news coverage, review aggregators (Yelp, Google, TripAdvisor), and structured local listing data.

For brands in high-competition markets like New York, the challenge is standing out in a category where hundreds of businesses carry similar authority signals. The GEO edge in local markets tends to come from three places: coverage in local publications with indexed digital editions, accurate and detailed Google Business Profile data, and review volume and recency on platforms AI engines retrieve from.

One thing local brands often miss: AI engines are increasingly willing to cite niche, high-credibility local sources over national ones for local queries. A mention in a well-regarded regional business publication can carry more GEO weight for a local query than a mention in a national outlet with no local context. That is an opening for brands that have invested in local PR and are not yet treating it as a GEO asset.

The entity layer matters here too. If your business has a Wikidata entry, a Wikipedia article, or listings in major industry directories that AI systems index, your visibility on local and category queries improves a lot. This is low-effort work next to the content production the other tactics demand.

What should a GEO audit cover in 2025?

A credible GEO audit has five parts.

First, a citation baseline: run 30 to 50 target queries across ChatGPT, Perplexity, Claude, and Google AI Overviews and record which sources appear. Do not start with your own brand name. Use the category questions your customers actually ask. This tells you where you stand before you touch anything.

Second, a content structure audit: for your top 20 to 30 pages, check whether each section has a direct answer in the first 40 to 60 words, whether factual claims carry inline citations or named sources, and whether each page has at least three extractable standalone factual sentences. Pages that fail these checks are your highest-priority rewrites.

Third, a schema audit: check which pages have FAQ schema, HowTo schema, Organization schema, and Article schema. Cross-reference against which pages you want cited by AI engines. Any gap is a technical fix, not a content project.

Fourth, an entity audit: verify that your brand has accurate, complete information in Google Knowledge Graph, Wikidata if applicable, your LinkedIn company page, Crunchbase if relevant, and major industry directories. These are the sources AI systems pull from for training data and for real-time entity resolution.

Fifth, a competitor gap analysis: for the queries where competitors appear in AI citations and you do not, document what those cited pages have that yours lack. The answer is almost always one of: more inline citations, cleaner direct answers, more structured schema, or coverage in a higher-authority publication.

Spawned's audit process covers all five layers and produces a prioritized action list, which is the most useful output if you are deciding where to spend limited time.

For tools that can automate the citation baseline step, brandrank.ai visibility insights analysis and ai visibility tool are both worth evaluating.

Sources

1. Aggarwal et al., Princeton / IIT Delhi / Georgia Tech, 'Generative Engine Optimization' (arXiv 2311.09735)
2. University of Amsterdam, RAG-based retrieval and content structure research (arXiv 2024)
3. Google, AI Overviews Help page
4. SE Ranking, AI Overviews Impact Study 2025
5. Perplexity AI, company blog / press releases 2024
6. Google Search Central, How Google Search Works: E-E-A-T and structured data
7. BrightEdge, AI Search Impact Report 2025
8. Ahrefs, AI Overviews frequency study 2024-2025
9. OpenAI, ChatGPT Search feature announcement 2024
10. Google Search Central, Introduction to structured data