GEO: How to Get Your Brand Cited by ChatGPT and Perplexity

Ask ChatGPT, Perplexity or Google which agency, tool or supplier to pick, and you get an answer, not a list of ten links. A handful of sources are cited inside that answer. If your brand is one of them, you are part of the decision. If it is not, the buyer may never see you at all.

Generative engine optimization, or GEO, is the work of earning those citations. This guide covers what the research actually shows, what the platforms say in their own documentation, and a practical way to measure progress, without the hype that has grown around the term.

What GEO is, and what it is not

The term comes from a 2023 paper by researchers at Princeton University and IIT Delhi, published at the KDD 2024 conference. It defines generative engines as systems that retrieve sources from the web and use a language model to write an answer grounded in them, with inline citations (Aggarwal et al., GEO: Generative Engine Optimization). GEO is the practice of shaping content so that these engines are more likely to use and cite it.

GEO is not a separate discipline that replaces SEO. Most answer engines retrieve pages through a search index before they write anything, so a page that is hard to crawl or poorly indexed starts at a disadvantage. Think of GEO as a layer on top of good SEO: same foundations, different finish.

Google is explicit on this point. Its guidance for AI Overviews and AI Mode says the same foundational SEO best practices apply, that a page must be indexed and eligible to show with a snippet, and that there are no additional technical requirements (Google Search Central). Be skeptical of anyone selling a secret technical trick for Google's AI features.

Why it matters now

The shift in behavior is measurable. Pew Research Center tracked the browsing of 900 US adults in March 2025. When a Google search showed an AI summary, users clicked a traditional result in 8% of visits, against 15% when there was no summary. They clicked a link inside the summary itself in just 1% of visits (Pew Research Center). Pew also found that 60% of searches starting with a question word, such as who, what or why, produced an AI summary.

Two conclusions follow. Fewer clicks reach websites from informational queries, and being named inside the answer becomes a visibility goal of its own, even when no click follows. A buyer who sees your brand cited three times while researching a problem arrives at the shortlist already knowing your name.

Analysts saw this coming. In February 2024, Gartner predicted that traditional search engine volume would drop 25% by 2026 as AI chatbots and virtual agents take over queries, and advised companies to focus on unique, useful content that shows expertise, experience, authoritativeness and trustworthiness (Gartner). Treat the figure as a forecast, not a measurement, but the direction matches what Pew observed.

What the research says gets content cited

The GEO study is still the most useful evidence because it tested specific changes to content and measured the effect. The researchers built a benchmark of 10,000 queries across many domains, rewrote source pages using nine different methods, and measured how much of each page made it into the generated answer.

The findings worth acting on:

• Evidence wins. in the benchmark, adding citations to credible sources, quotations from relevant experts and statistics boosted source visibility by over 40% across various queries.
• Keyword stuffing fails. the classic SEO habit of repeating the target phrase gave little to no improvement in generative engine responses.
• Clarity helps. improving fluency and making text easier to understand also lifted visibility, though less than adding evidence.
• It works on real engines. tested on Perplexity, the methods improved visibility by up to 37%.
• Domain matters. the best method varied by subject, so test on your own topics rather than copying a universal checklist.

One result matters most for smaller brands. In the benchmark, lower-ranked websites benefited far more than top-ranked ones. Adding citations raised visibility by 115.1% for sites ranked fifth in classic search results, while the top-ranked site lost 30.3% on average. In classic SEO, page one position five rarely beats position one. In an AI answer, a better-evidenced page can.

Step 1: Make sure AI engines can reach you

Before any content work, check that the crawlers that feed AI search are allowed in. This is the most common silent failure we see: a security plugin, CDN rule or blanket AI block that shuts out the bots a brand actually wants.

The key distinction is between search crawlers and training crawlers. OpenAI runs them separately. OAI-SearchBot surfaces websites in ChatGPT search, and sites that block it will not be shown in ChatGPT search answers. GPTBot collects content that may be used to train OpenAI's models. Each setting is independent, so you can allow search and refuse training (OpenAI crawler documentation).

Perplexity follows the same pattern. PerplexityBot is designed to surface and link websites in Perplexity results and is not used to crawl content for AI foundation models, and Perplexity recommends allowing it in robots.txt (Perplexity crawler documentation).

A minimal checklist:

• Allow Googlebot and Bingbot. they feed Google AI features and Microsoft Copilot.
• Allow OAI-SearchBot. for ChatGPT search. Decide on GPTBot separately, based on your view on training.
• Allow PerplexityBot. for Perplexity answers.
• Write it explicitly. in robots.txt, a block of User-agent: OAI-SearchBot followed by Allow: / is enough, repeated for PerplexityBot. Add User-agent: GPTBot with Disallow: / only if you want to refuse training.
• Check the CDN and firewall too. Google notes that crawling must be allowed in robots.txt and by any CDN or hosting infrastructure. A robots.txt that says yes means nothing if the firewall says no.
• Keep key content in text. answers are written from text. Pricing, specs and definitions locked inside images or scripts are easy to miss.

Step 2: Write pages an engine can lift

An answer engine pulls passages, not whole pages. The easier it is to lift a clean, self-contained passage that answers the question, the more likely your page is used.

1. Lead with the answer. put a two to four sentence direct answer at the top of the page, before context or story. If the question is "what is GEO", the first paragraph should define GEO.
2. Use headings that match real questions. write H2s the way a buyer would phrase the query. "How much does a fractional CMO cost" beats "Our pricing philosophy".
3. Make lists and tables extractable. steps, comparisons and criteria in lists or tables are easier to reuse accurately than the same facts buried in prose.
4. Add an FAQ that answers follow-ups. Google says AI Mode uses a query fan-out technique, running related searches across subtopics. A good FAQ covers those adjacent questions on the same page.
5. Name things precisely. use the full names of products, standards, places and organizations. Vague references are hard to match to a query.

None of this means writing for machines. Pages that are clear for an engine are almost always clearer for people. For the writing itself, our guide to AI content marketing without sounding generic covers how to keep a distinct voice.

Step 3: Back every claim with evidence

This is where the research points hardest. Engines are trying to give a reliable answer, and pages that show their evidence are easier to trust and quote.

• Cite primary sources. link the original study, official documentation or regulator page, not a blog that summarised it.
• Use specific numbers. a figure with a source, a sample size and a date is more quotable than an adjective. "Clicked in 8% of visits" beats "far fewer clicks".
• Quote named experts. a short, attributed quote from a recognised person or organisation was one of the strongest methods in the GEO study.
• Publish what only you know. your own benchmarks, client results and pricing data are content no one else can supply. That is what makes a page worth citing over a competitor's.

One warning: never invent or round up a statistic to make a page more quotable. Answer engines spread errors fast, and a wrong number tied to your brand name is worse than no number.

Step 4: Be cited off your own site

Your website is only one input. Pew found that Wikipedia, YouTube and Reddit were the most frequently cited sources in Google's AI summaries, together accounting for 15% of the sources listed (Pew Research Center). Engines build their view of a brand from many places.

So the off-site work matters:

• Keep facts consistent. your name, category, location and offer should read the same on your site, Google Business Profile, LinkedIn, directories and review platforms.
• Earn mentions in trusted places. industry press, partner pages, podcasts and analyst lists all add independent signals about what you do.
• Show up where buyers ask. genuine, expert answers in relevant communities and video explainers on your core topics feed the same sources answer engines read.

How to measure AI visibility

Measurement is still immature, but there is now enough to track progress rather than guess.

Bing Webmaster Tools added an AI Performance report in public preview in February 2026. It shows total citations of your content across Microsoft Copilot, AI summaries in Bing and select partner integrations, which URLs are cited, and a sample of the grounding queries the AI used to retrieve your pages (Bing Webmaster Blog). It is the closest thing to a citation report any major engine offers today.

For ChatGPT, OpenAI says it automatically adds utm_source=chatgpt.com to referral URLs from ChatGPT search results, so that traffic can be tracked in Google Analytics and similar tools (OpenAI publishers FAQ). Build a channel group or saved report for AI referrers, including chatgpt.com and perplexity.ai.

Google does not separate AI features in its reporting. Traffic from AI Overviews and AI Mode is counted inside the Web search type in Search Console's Performance report (Google Search Central). Watch impressions against clicks on question-style queries for signs of the shift.

Then add a manual check. Write 20 to 30 prompts a real buyer would ask about your category, run them monthly in ChatGPT, Perplexity, Gemini and Copilot, and record whether your brand is mentioned, whether it is cited with a link, and which competitors appear. It is crude, but it is the only way to see the answer your buyer sees.

If you already run AI on your marketing data, our guide to AI marketing analytics shows how to fold these signals into one view.

What to stop doing

• Keyword stuffing. the GEO study found it gave little to no improvement in generative answers.
• Mass-produced AI pages. Gartner's advice was the opposite: unique, useful content with real expertise. Thin pages give an engine nothing worth quoting.
• Blocking every AI bot by default. decide bot by bot. Blocking search crawlers removes you from the answers your buyers read.
• Chasing silver bullets. new files, tags and "AI ranking hacks" appear every month. Google states there are no additional technical requirements for its AI features. Spend the time on content and evidence first.

A 30-day GEO starting plan

1. Week 1: access and baseline. audit robots.txt and CDN rules for the AI search crawlers, verify your site in Bing Webmaster Tools and Search Console, set up AI referrer tracking, and run your prompt panel once to record a baseline.
2. Week 2: pick the pages that matter. choose the five to ten pages that answer the questions closest to a buying decision: service pages, pricing, comparisons and your best guides.
3. Week 3: rewrite for extraction and evidence. add an answer-first summary, question-style headings, a real FAQ, and primary-source citations with specific numbers to each page.
4. Week 4: off-site and review. fix inconsistent brand facts across profiles and directories, pitch one piece of original data to a trusted industry outlet, and rerun the prompt panel to compare.

GEO rewards the same thing good marketing always has: being the clearest, best-evidenced answer to a question your buyer actually asks. The difference is that the reader checking your evidence is now a machine, and it is less forgiving of vague claims than a human skimming a page.

If you want an outside view on how your brand shows up in AI answers today, and what to fix first, that is the kind of work we do at Gaveau Strategy.