I speak to a lot of heads of marketing, and many of them have built the same machine. An SEO agency, a content team, and an article for every long-tail keyword they could find. The machine still runs. Fewer people click.

Buyers now ask ChatGPT, Gemini or Claude and take the answer. That answer doesn't rank ten pages. It cites a handful of sources, and most of them aren't the brand's own website. Most teams are still tuning their pages for a list of links buyers no longer read.

Most of what follows comes from my conversation on the GTM & AI Operators Show with Nitya Chidambaram, co-founder of HeyAmos, which helps brands show up in AI search, who previously scaled ShopBack across 13 markets and worked at Amazon and Practo, and a discussion with Nupura Ughade, Content, SEO and Brand Lead at HyperVerge, who previously led content at SenseHQ and Chargebee.

You can watch my conversation with Nitya Chidambaram on AI search here.

AEO and GEO: what the two names mean

You'll hear two names for this work, often in the same sentence.

AEO, or answer engine optimization, is the older term. It has been around since search engines and voice assistants started answering questions directly instead of only listing links. The goal is to be the source behind the answer, not one link among ten.

GEO, or generative engine optimization, comes from a 2023 research paper by researchers at Princeton and IIT Delhi, presented at the KDD conference in 2024. It looks at tools that write a fresh answer from many sources, such as ChatGPT, Perplexity and Google's AI Overviews, and at how content can show up in what they write. The authors found that some changes to content raised a source's visibility in those answers by up to 40%.

In practice, most teams use the two words for the same job: getting your brand named and cited when someone asks an AI tool. If there's a difference, it's where you start. AEO starts from the question a buyer asks. GEO starts from how the AI builds its answer. I use AEO in this post, but everything here applies to both.

If you're not in the answer, you don't exist

When Nitya started her company in Singapore, she needed a business bank account. The old way would have meant searching Google, scrolling past sponsored ads, reading reviews, comparing options and asking a friend. She wasn't willing to spend more than 30 minutes on the decision, and every bank looked roughly the same anyway.

So she gave ChatGPT three banks and her context. Three founders, a fresh pre-seed round, plans to go global, so exchange rates mattered, and she needed Stripe integration. It named one bank. She checked the details, applied online and was done.

Now imagine you're a bank that offers exactly the same thing but wasn't on her list. In the old journey, she would at least have seen your link before skipping it, and you could have measured that. In this one, you were never in the room. You don't lose the comparison. You never enter it.

The data shows the same shift at scale. When Pew Research tracked 900 US adults through 68,879 Google searches in March 2025, people clicked a result 8% of the time when an AI summary appeared, against 15% when there wasn't one. Only 1% clicked a link inside the summary itself. Across Google, Similarweb found that searches ending without a click rose from 56% to 69% in the year after AI Overviews launched.

That's why so many brands now see sales arrive without a clear trail. The buyer did the research. It just didn't happen on their website.

Buyers used to skip you. Now, if the AI doesn't name you, they never see you at all.

How AI search actually chooses

A person searching Google reads the first page, opens a few of the ten links and picks one. An AI tool works at a different scale. It pulls from hundreds or thousands of sources at once, some from what it learned in training and some from live searches it runs for that question, then writes one answer from them.

It also searches differently from a person. Google calls this query fan-out: AI Mode breaks one question into many related searches, including ones the user never typed, and merges the results. Ask for the best boutique hotels in Tokyo, and behind the scenes that becomes a set of long, specific queries, often with the current year in them.

That last part matters more than it sounds. AI tools lean towards recent sources. In Muck Rack's analysis of more than 25 million links cited by ChatGPT, Claude and Gemini, more than half of the journalism cited had been published in the past 12 months. A page nobody has touched since 2023 is competing with one hand tied.

One question becomes many searches, hundreds of sources and a single answer.

Good SEO is still the starting line

It's tempting to treat AI search as a clean break from SEO. The evidence says otherwise. Before Nupura's team had any AI search strategy at all, they were already getting cited. The pages AI picked were the ones that were already strong for SEO, on topics where they had real depth.

"It requires different tactics, but good SEO is not optional."

Nupura Ughade, Content, SEO and Brand Lead, HyperVerge

What changes is the bar. You don't need a top-10 ranking to be cited, which levels the field for smaller brands. A useful rule of thumb: ranking somewhere in the top 30 usually means the basics are in place: good pages, relevant topics, sound data. Pages like that tend to show up in AI answers too.

Many sites never reach that line. Give ChatGPT the address of a site made with an AI website builder and ask what the company does, and it may tell you it sees an empty HTML file. That's not a glitch. In late 2024, Vercel and MERJ found that none of the major AI crawlers, including OpenAI's and Anthropic's, run JavaScript. If your content only appears once a browser builds the page, AI tools see nothing.

The starting line is simple, and most of it is old SEO hygiene:

• Pages that load their content without needing JavaScript

• No accidental blocks on AI crawlers

• Real depth on the topics you want to be known for

• Page titles that match how buyers actually phrase the question

Good SEO is the entry ticket to AI search, not the finish line.

Your website is a small share of the answer

Here's the part most teams haven't absorbed. Even a perfect website is only a small part of what AI cites. In travel, HeyAmos's data shows that only 16% of citations come from the brand's own site. About 65% come from high-authority sources: news coverage, review sites, third-party blogs, forums and Wikipedia. In some B2B categories, the brand's own share drops as low as 6%.

Independent research points the same way, and often further. Shero Commerce studied 60 product categories across Google AI Mode, ChatGPT and Perplexity. Brand-owned pages made up just 2.8% of the sources cited. Even when an AI tool recommended a brand, it cited that brand's own page only 31% of the time.

There's one clear exception. In e-commerce and marketplaces, the product page is the source of truth for price and specs, so the brand's own site matters a lot more. Everywhere else, there's a ceiling.

"You can publish 10,000 blogs and you won't see any impact."

Nitya Chidambaram, co-founder, HeyAmos

That doesn't make the website irrelevant. Nupura currently puts about 60% of her effort into earning mentions elsewhere and 40% into her own site. But she's clear about why that split works for her: the site already has a solid SEO foundation and no thin pages. Third-party mentions point back to your site, and AI tools read it too. If the pages they land on are vague, the mentions are wasted. A new or thin site should fix that first.

Your own website is a small share of what AI cites, in every study I looked at.

Earned media is the new moat

If most citations come from other people's pages, the question becomes how you get onto those pages. The strongest answer is earned media: coverage someone else chose to write about you. Muck Rack's research found that 84% of AI citations come from earned media, a share that has held between 82% and 89% since July 2025. Paid content made up 0.3%.

HeyAmos runs its own experiments on which kinds of coverage move AI visibility most. On their numbers, one published news article about you makes you about 228% more likely to be cited than your own website does. They saw it first-hand: when the startup news site Tech in Asia covered HeyAmos, its visibility jumped. A publication like that has done its own checks before writing about you, and AI tools seem to trust that.

The catch is that someone else has to choose you. A press release on a newswire only counts if a journalist picks it up. Nupura has tried both paid media placements and steady, organic PR, and organic has always done more for citations. In her words, quality matters more than quantity here. That lines up with Muck Rack's 0.3%.

The other high-authority sources each have their own rules:

• Review sites. G2 is one of the most cited sites for software. But G2's own study of about 30,000 citations found that 10% more reviews meant only about 2% more citations. Reviews help, but piling them up doesn't move much.

• Wikipedia. Very high authority, and very hard to get into. Its editors set a high bar for what they accept.

• Reddit. Once brands saw AI quoting Reddit, many started seeding comments there. ChatGPT's citations of Reddit dropped sharply in late 2025, though the cause is disputed. Either way, it's a volatile place to build on.

All of this changes what PR is for. It used to sit at the top of the funnel as a brand exercise. Now it's one of the most direct ways to show up when buyers ask AI for a shortlist.

The harder a mention is to earn, the more weight AI gives it.

What works, and what's a waste of time

The first wave of AI search advice was a checklist: add schema, write FAQs, refresh old pages, publish comparison pages, get onto Reddit. Some of it works. A lot of it is busywork. Here's how it has played out at HyperVerge.

What made a clear difference:

• Honest comparison pages. Pages written to help a buyer choose, not to win the argument.

• Original data nobody else has. Benchmarks and industry research, sent out through PR so other sites cover it.

• Basic technical hygiene. For example, making sure you aren't blocking AI crawlers by accident.

What didn't move much:

• Schema markup

• llms.txt files

• Content refreshes that change the date but not the substance

• Seeding Reddit

Two of these sit against what Nitya said on the show, and the gap is useful. She counts schema as part of the starting line that makes a site readable. Nupura saw no real change from it. My read: schema helps AI tools parse a page, but it won't make anyone cite a page that isn't worth citing.

The second is comparison pages. Nitya's view is that a brand writing its own "us vs them" page no longer works, because anyone can write one. Nupura says honest comparison pages are among the things that work best. The difference is the word honest. A page built to win the comparison reads like an ad, to people and to AI. A page that genuinely helps a buyer choose, even when the answer isn't you, reads like a source.

The same logic covers refreshing. AI does favour recent pages, but changing a date isn't the same as having something new to say. Update a page when the facts have changed, and AI has a reason to treat it as current.

Substance gets cited. Tactics built to game AI mostly don't.

Measuring what you can't click-track

SEO came with a clean trail: page views, sessions, last click, the exact path a visitor took. AI search breaks that. A buyer can read your name in a ChatGPT answer, never visit your site, and book a demo two weeks later. Nitya expects session-based attribution to collapse, and the whole foundation for tracking it will need rebuilding.

Until that exists, the practical answer is proxies. Nupura uses three:

1. Self-reported attribution. An open-text "How did you hear about us?" field on the demo form. People do write in ChatGPT or Gemini, and from there she can follow those leads through the pipeline.

2. Referral traffic from AI tools. Some buyers do click through, and that traffic can be tracked by source.

3. Directional signals. How often the brand is mentioned when you run the prompts buyers are likely to ask, and whether branded search is rising.

The harder part is what to tell leadership. Her approach is to be straight with them: there isn't full visibility yet into how AI search affects pipeline, but here are the signals showing what's working. So many leads cited AI this quarter, the brand's mention rate on buyer prompts moved from one number to another, and branded search is up.

She also owns PR and brand, which helps. Both have always been measured this way, by direction rather than precise attribution. AI search simply joins them. Most performance marketers will find that uncomfortable. They'll have to get used to it.

When the trail disappears, measure the direction of travel.

When agents do the shortlisting

The next step is already visible. Buyers aren't only asking AI for advice. They're starting to let AI act for them. In September 2025, Stripe and OpenAI launched Instant Checkout in ChatGPT and the Agentic Commerce Protocol, so people can buy from some merchants without leaving the chat. The payment rails for agents are being laid now.

Nitya's picture of B2B is blunt. An agent will search, decide that a particular CRM is right for your stage of company, and go buy it. If you aren't in the set it considers, you never hear about the deal.

Nupura sees half of that already. Agents helping buyers draw up a shortlist is happening today. Agents actually making the purchase is unlikely in her market, which sells to banks, insurers and non-bank lenders. Companies selling self-serve products may see it sooner.

Whatever the pace, the job lands in the same place. Whether it's a person asking ChatGPT for three options or an agent with a budget, the shortlist gets written from what credible sources say about you. Being on it is the whole game. I've written separately about how B2B buyers build that shortlist, and what it means for your brand.

Websites were built for a buyer who read ten links and picked one. The next buyer reads a thousand sources in seconds and names three. Most teams are still polishing the page. The ones that get named are making sure someone else is writing about them.

FAQs

What is answer engine optimization (AEO)?

AEO is the work of making a brand show up in answers from AI tools such as ChatGPT, Gemini, Claude, Perplexity and Google's AI Overviews. Where SEO aims to rank a page in a list of links, AEO aims to have your brand named and cited inside the answer itself. Generative engine optimization (GEO), a term from a 2023 research paper, describes the same goal from the AI's side, and most teams use the two names interchangeably.

Is AEO different from SEO?

The foundations are the same. AI tools tend to cite pages that are already strong for SEO: easy to crawl, deep on a topic and clearly titled. The difference is what comes after. A top-10 ranking matters less, and mentions on third-party sites such as news, review platforms and Wikipedia matter much more.

How does ChatGPT decide which sources to cite?

AI tools combine what they learned in training with live searches run for each question. They often split one question into many related searches, then draw on hundreds of sources to write a single answer. They favour recent, credible and clearly relevant pages, and lean heavily on earned media rather than brand-owned sites.

How do I get my brand cited in AI search results?

Start with a crawlable site that has real depth on your topics. Then earn coverage in credible publications, publish original data that others will reference, keep comparison pages honest and useful, and update pages when the facts change. Independent research suggests most citations come from earned media, so PR has become one of the most direct levers.

Does schema markup help with AI search visibility?

The evidence is mixed. Schema can help AI tools read a page's structure, and some practitioners treat it as basic hygiene. Others report no measurable change in citations from schema alone. It's worth having, but it won't make an AI tool cite a page that isn't worth citing.

How do you measure traffic and pipeline from AI search?

Most AI-influenced buyers never click through, so standard attribution misses them. Teams use proxies: an open-text "How did you hear about us?" field on forms, referral traffic from AI tools, how often the brand appears when running typical buyer prompts, and trends in branded search.

Will AI agents buy B2B software on their own?

Agents already help buyers build shortlists. Agents completing purchases is starting in consumer commerce, with checkout built into ChatGPT, but in most B2B markets it's still early, especially in regulated industries. Self-serve products are likely to see it first.