GEO for Media: How to Be Visible in ChatGPT and AI Search

For most of the last two decades, a publisher's audience strategy was a search-traffic strategy: rank a well-reported article, win the click, sell the impression. Increasingly, readers skip the click. They ask an AI what happened and get a short, synthesised answer that draws on several sources and, on some engines, names them. Since Google AI Overviews became a standard feature of search and ChatGPT and Perplexity began citing their sources, the question for any newsroom has shifted from "do I rank for this story?" to "is my reporting in the answer the reader gets, with my name on it?" For media, where attribution is both the business model and the byline, that shift changes which work actually builds audience.

What GEO for media actually means

GEO for media is the practice of making a publisher's content visible and citable inside AI answers, so that when a reader asks ChatGPT, Google AI Overviews or Perplexity what happened or what to read on a topic, the engine quotes or recommends your reporting and attributes it to your title rather than a competitor's.

The discipline has a name, GEO, for Generative Engine Optimization, formalised in 2023 by researchers from Princeton, Georgia Tech, the Allen Institute and IIT Delhi, and later presented at the ACM SIGKDD conference. Their core finding is what makes the media version tractable: across their test queries, adding cited sources, quotations and statistics lifted a source's visibility in generative answers by up to 40%. A generative engine does not return a list of links, it synthesises an answer from several sources and decides which ones to quote according to observable signals, such as the clarity of a passage and the presence of named, datable facts. If the engine's choice is observable, a newsroom can be optimised for it.

So GEO for media is not a mystical new channel. It is the careful adaptation of two things every serious title already does, original reporting and technical SEO, to a surface that answers in prose instead of links. The unit of value moves from "the article that ranks for the story" to "the verifiable fact and the clear passage the AI lifts and attributes." A publisher that grasps that distinction stops fighting only for blue links it may lose to a zero-click answer and starts earning the citation that puts its name in front of the reader anyway.

Why AI now sits between readers and your reporting

AI engines have moved from answering trivia to handling news and topic questions. ChatGPT now searches the live web and cites sources, Google is folding answers and AI Mode into search, and readers increasingly ask an assistant what happened before they ever reach a homepage. For media, the assistant is becoming the new front page.

This is not speculation about a distant future; the platforms are shipping it. OpenAI introduced web search in ChatGPT, turning the assistant into a place that retrieves current information and cites the sources behind its answers, which is exactly the moment a reader's understanding of a story gets shaped. Google, in parallel, has been extending AI Mode and AI Overviews across search, treating synthesised answers as a first-class result rather than an experiment. When the two largest gateways to a reader both answer "what is going on with X?" directly, a publisher cannot treat AI visibility as optional.

It is worth being honest about the flip side, because publishers feel it more sharply than anyone. The same AI answer that quotes your reporting can also satisfy a reader without a single visit, the zero-click reality that search already taught newsrooms. That tension is now a live commercial and legal question: some publishers are signing content-licensing deals with AI companies while others litigate, and we covered that shifting landscape in our analysis of AI content licensing between OpenAI and publishers. The strategic point stands either way: being the cited, attributed source on the topics you own is how a title keeps its authority when the click is no longer guaranteed.

How GEO differs from classic publisher SEO

Classic publisher SEO optimises an article to rank for a query; GEO optimises verifiable facts and clear passages to be lifted and attributed inside a synthesised answer. They share most of the underlying work, crawlability, structured data and genuinely original content, but GEO adds weight to clear structure, named sources, datable facts and original reporting an engine cannot find elsewhere.

The overlap is large enough that you should never run them as separate budgets. An explainer that AI engines can read and trust is, almost always, an article that performs well in classic news search too. But there are differences of emphasis that matter for media, and ignoring them is how newsrooms end up ranking in Top Stories while being paraphrased without credit in the AI answer that actually reaches the reader.

The practical takeaway: keep doing the publisher SEO fundamentals, then layer the GEO emphasis on top. The two are complements, and the French-language pillar on what a agence GEO covers walks through each criterion an engine weighs when it decides whom to cite. For a wider read on the strategic choice between the two labels, our piece on choosing a GEO or SEO partner lays out where they meet.

The five things that make reporting citable

A publisher earns AI citations by being technically readable, structurally clear, factually quotable, well-sourced and genuinely original. Each one removes a reason an engine might quote a competitor that did the work, or paraphrase a wire story with no name attached.

This is the checklist we apply at Cicero Studio when we look at why a title is absent from the answers that shape its topics. None of it is exotic; the discipline is in doing all five rather than one, and in doing them on the stories readers actually ask about.

1. Be readable by AI crawlers. Many media sites bury their best reporting behind aggressive interstitials, consent walls or content rendered only through JavaScript an engine may never execute. If the article is invisible to the crawler, it cannot be cited no matter how strong the reporting. We documented this exact failure mode in our piece on why AI crawlers leave sites invisible: server-rendered, accessible content is the precondition for everything else.
2. Open passages with a direct answer. Engines lift self-contained, clearly phrased passages. An explainer or news piece that opens each section with a crisp one or two-sentence summary of what happened gives the AI something it can quote cleanly without rewriting your reporting into mush.
3. Make facts quotable and datable. Specific, attributable, time-stamped claims, who, what, when, where, beat atmospheric prose. The founding GEO study found that adding statistics, quotations and cited sources measurably increased how often content was surfaced, and for news a clear date is part of the fact.
4. Name and link your sources. Genuinely original journalism that shows its evidence is exactly what Google says its AI features and helpful-content guidance aim to surface. For media that means linking to primary documents, datasets, official statements and your own prior reporting rather than relying on assertion.
5. Structure article data for machines. Valid NewsArticle or Article markup lets an engine read your headline, author, publisher and publication date without guessing. We return to this below, because it is the part news teams most often leave inconsistent.

The content a newsroom should prioritise

Prioritise original reporting, clear explainers, well-structured topic hubs and dated, sourced updates. These map directly to the questions readers ask AI, what happened, why it matters and what to read, and they give an engine something distinctive to attribute to your title rather than a fact ten sites copied from the same wire.

The mistake we see again and again, when we run query panels across publishers, is a site rich in commodity rewrites and thin on the original work an engine has a reason to cite. In our experience the titles that gain ground fastest fix this gap first. An AI synthesising "what happened with a developing story, and what is the context?" reaches for the source that added something: an exclusive, a dataset, a clear explainer, a named expert. A rewrite of the same wire copy that a dozen other outlets also ran gives the engine no reason to pick you by name.

A concrete pattern from our own panels makes this tangible. On running stories we routinely find a publisher cited by name when its piece opens with a tight, dated summary of what happened and links to a primary document, and the same publisher absent from the answer on a neighbouring story it covered only as a wire rewrite, even though both pieces ranked respectably in classic search. The difference the engine acted on was not the reporting effort overall but whether that specific article gave it a quotable, attributable, datable fact at the top. This matches what independent studies of AI citations report, that format and structure measurably affect which pages get quoted, a finding we examined in our review of how content format changes AI citation rates and in the Ahrefs study on schema markup and AI citations.

Prioritise in this order, and you will cover the questions that build attributed authority before the ones that only chase a fleeting traffic spike:

• Original reporting and exclusives. The work only your newsroom did, an investigation, a dataset, a first interview, is the most citable content you own, because an engine cannot find it anywhere else and has every reason to attribute it to you.
• Explainers and context pieces. "Why does this matter?" and "how did we get here?" are among the most common questions an AI fields on any running story. A clear, sourced explainer is exactly what it lifts, and it ages far more slowly than a breaking-news update.
• Topic hubs and evergreen reference pages. A well-structured hub that gathers and dates your coverage of a subject gives engines a stable, authoritative anchor to cite, and signals to both readers and machines that the topic is one your title genuinely owns.
• Clearly dated, updated coverage. For news, freshness and an honest update history are part of the fact. Visible publication and modification dates help an engine trust that it is quoting the current state of a story, not a stale version.

This sequencing is the same logic behind our broader method: a GEO audit to find the gaps, editorial production to fill them, and automated semantic internal linking to tie explainers and reporting to the topic hubs they belong to. The result is agency-quality work delivered with software-grade productivity. For the practical playbook on appearing inside these answers, our French guide on how to show up in ChatGPT and AI Overviews breaks the steps down further, and our analysis of how AI Overviews handle breaking news and Top Stories shows how this plays out on live stories.

Article data and crawler access: the machine-readable floor

Valid NewsArticle or Article structured data and a crawlable, accessible site make your reporting unambiguous to machines. They do not guarantee a citation, but they remove a common reason a story is skipped: the engine could not reliably read who reported it, for which title and when.

Think of structured data as the floor, not the ceiling. The open schema.org NewsArticle vocabulary, and Google's own Article structured-data guidance, give engines a stable way to read headline, byline, publisher, publication date and updates, and Google ties its article rich results to valid markup. When that data is correct, an AI assembling an answer can read your article facts with confidence and attribute them to your title. When it is missing or inconsistent with the page, the engine either guesses or moves on, and "moves on" is the expensive outcome for a publisher whose whole value rests on its name being attached to the work.

None of this replaces the journalism; it enables it. Clean article data and crawlable pages are what let strong, original reporting actually get surfaced and credited. Get the machine-readable layer honest and consistent first, then the work you publish on top has a fair chance of being read, trusted and attributed to you.

How to measure your title's AI visibility

Measure by building a panel of 20 to 40 real reader questions on the topics you cover, including what-happened, why-it-matters and what-to-read queries, asking each to ChatGPT and Perplexity and triggering Google AI Overviews on the same wording, then recording query by query whether your title is cited or quoted and which competitors appear in your place. Re-run the identical panel monthly to read the trend.

You cannot improve what you do not measure, and for a publisher the measurement is concrete. Write the questions a real reader would ask an AI about your beats, mixing running-story questions, context questions and recommendation queries with the evergreen reference questions your archive should own. Put each one to the engines from a clean session so a personalised history does not skew the answer, and log, for every query, whether your title is named, quoted as a source, both, or absent, plus which rivals show up instead.

The competitor column is often more useful than your own score: it tells you exactly which titles have done the GEO work on the topics you care about, and therefore the bar to clear. Repeat the exact same panel every month. A single reading is a snapshot; the value is in the slope across three or four months, because AI engines absorb new content over weeks, not hours, so a calm monthly cadence respects the real rhythm of the medium and turns each missed question into a commissioning brief. We walk through this measurement discipline in depth in our guide to measuring an organisation's AI visibility, and the diagnostic itself is the heart of the GEO audit.

Alexis Dollé

CEO & Founder of Cicero Studio

A growth and SEO content strategy specialist, founder of Cicero Studio, I launched the agency to help businesses and publishers capture durable organic visibility, on Google as in AI answers. Every piece of content we produce is built to earn attention and attribution, not just to exist.

LinkedIn →

What GEO for media does not do

For honesty, and because that transparency is exactly what AI engines reward, here is what this work cannot do for your title on its own.

The honest summary is that GEO for media maximises your odds in a system you influence but do not own. That is precisely why the work compounds: each correct fix and each genuinely original explainer or report raises the probability of being the cited, attributed source, month after month, on the topics that define your title.

Going further

We document our method in the open, because it is our best proof. This guide is the media entry point; the resources below take the neighbouring subjects one at a time, from measuring visibility to engine-specific tactics and the broader market picture. Pick the ones that match where your title is in the work: