How to get cited in AI answers: my AEO plan

October 27, 202612 min read

ai-search / aeo

Gold signal rings radiating from a point on a dark field, cover image for "How to get cited in AI answers: my AEO plan"

To get your brand cited in AI answers from ChatGPT, Google AI Overviews, Gemini and Perplexity, start with access. The AI search crawlers have to be able to reach your site. Then write pages where every section opens with its answer and names real products and facts. Keep your best pages updated, and earn mentions on the third-party sites each AI trusts. Finally, track citations directly. Google rankings still matter, but they no longer decide the shortlist on their own.

Chapters in the video

Why AI citations are a discoverability problem

AI citations decide whether a buyer hears your name at all. When someone asks ChatGPT to recommend a brand like yours, ChatGPT returns a short list of options with a reason next to each one. If your brand is not on that list, the buyer is unlikely to go looking through ten blue links to find you.

That changes the problem you are solving. In the video I put it this way: "But now, if you aren't being cited by AI, you have a discoverability problem." Losing a Google ranking used to mean less traffic. Missing from an AI answer means the buyer may never learn you exist.

To build this plan, I worked through the Ahrefs AEO course, about 30 other sources, and the official documentation from Google and OpenAI. I also built a tracker that asks ChatGPT, Perplexity, Gemini, Claude and Google the same buyer questions, so I can compare how each one picks its sources.

How AI search decides who to cite

AI search picks sources in three stages: it retrieves live results, it splits your question into many smaller searches, and it then favors sources that agree with each other and are recent.

The first thing to understand is where the information comes from. AI models draw on two places:

Training data. This is a snapshot of the internet that gets built into the model. It does not update on a set schedule, and a page you publish today may never make it in.
Live search. The AI searches the web while it answers. This is called retrieval augmented generation, or RAG. Most of your opportunity is here.

The second thing is the query fan-out. ChatGPT does not search your exact question word for word. It breaks the question into a set of sub-searches and stitches the results into one reply. A question about planning a five-day trip to Japan might turn into searches on Tokyo neighborhoods, November weather in Tokyo, and whether the rail pass is worth the money.

Most of these fan-out searches are phrases real people rarely type. "So don't treat fan-out searches like a new keyword list." Write the way you would explain the topic to a customer out loud, broken into clear sections that each answer one part of the question.

The third thing is trust, and two signals drive it:

Consensus. If many sources say the same thing about your brand, the AI is more likely to repeat it.
Freshness. AI assistants favor recently updated pages, so a stale page is easier to pass over.

Ranking in Google's top 10 also does not guarantee a citation in Google AI Overviews. SEO is still the foundation. It is no longer the whole house.

Each AI platform cites different sites

ChatGPT, Google AI Overviews, Perplexity and Google AI Mode pull from different sources, so a single strategy for all of them leaves visibility on the table. The sites each platform cites most often overlap very little.

Here is how the platforms differ, based on my research for this plan:

Google AI Overviews lean on large, trusted sites, and on YouTube most of all.
ChatGPT leans toward publishers and big community sites such as Reddit, Wikipedia and Forbes. The pages ChatGPT cites tend to sit on very high-authority domains, and its sources rarely match Google's top 10.
Perplexity behaves the most like Google, with a larger share of citations coming from Google's top 10. If you already rank well, Perplexity is probably where you will see the fastest wins.
Google AI Mode is the surprise. AI Mode and AI Overviews come from the same company, yet the two share few sources even though their answers mean nearly the same thing.

Build your plan around the platform your buyers are most likely to use. A B2B buyer researching in ChatGPT and a local shopper reading AI Overviews need different offsite work from you.

What content gets cited by AI

AI cites content that is structured and specific, and length makes little difference. Longer pages do not earn more citations in Google AI Overviews, and plenty of cited pages are short.

Four habits matter more than length:

Lead with the answer. Put the main point in the first sentence of each section. Readers skim the top of a section, and AI models pay the most attention to the start of a passage too.
Make every section stand on its own. AI cuts your page into chunks, and you cannot control where the cuts land. Pull one section out and read it alone. If it does not make sense, rewrite it.
Be specific with names and facts. "This tool helps productivity" gives the model nothing to grab. "Ahrefs Keywords Explorer finds keywords with low difficulty and high traffic potential" does. A 2024 study called GEO tested ways of rewriting pages and found that adding statistics, quotes and sources boosted visibility in AI answers by up to 40% (Aggarwal et al., GEO: Generative Engine Optimization). That was a lab test where the AI was handed the pages, so it does not promise that AI will find yours.
Keep it fresh. This is the habit people sleep on most, because AI assistants favor recently updated pages. Find your sleeper pages before you write new ones. Those are pages that have backlinks and once drove traffic, or still do, but have plateaued or started to decline. Update them with new data, new examples and new metadata. Changing only the date does not work.

If you created a framework yourself, put your brand name on it everywhere it appears. If the AI cannot tell you created it, the AI may describe the idea and leave your name out.

How to earn mentions off your site

Your website alone will not get you cited, because AI weighs what the rest of the web says about you. I now put brand mentions across the web ahead of backlinks and domain rating when I plan offsite work.

Keep one caveat in mind. Big brands get mentioned more partly because they have bigger PR and social footprints, so mentions and AI visibility rise together for reasons that go beyond cause and effect. I do not read this as a reason to post everywhere. I read it as a reason to show up consistently in the communities where your target customer already spends time.

Most AI brand mentions carry no link, and unlinked mentions still help AI learn which brands belong to which topics. I sort the places to earn mentions into three tiers:

Tier one: editorial coverage. These are comparison posts, best-of lists and reviews on well-known sites. For "best" style questions in ChatGPT, list posts show up often among the cited pages. If you are on those lists, you are in the pool. That pattern applies to "best" questions, so your how-to guides do not need to become lists. Local businesses should add customer reviews and listing sites to this tier, since ChatGPT checks third-party sources such as the Better Business Bureau before recommending a business.
Tier two: community forums. These include Reddit, Quora and the niche forums in your vertical. Reddit is one of the sites ChatGPT cites most. Find real questions your product answers and give real answers. Do not spam.
Tier three: your own channels. These include YouTube, podcasts and LinkedIn. Every place your brand shows up in the right context adds one more signal.

YouTube is my favorite channel, and I see YouTube mentions as a strong signal for ChatGPT visibility. The evidence I have read comes from big brands, so treat YouTube as a test worth running. If you make videos, put the search phrase in the title and mention it naturally throughout, without repeating it at the viewer.

One experiment from the Ahrefs course shows why you need to state the facts about your brand yourself. Ahrefs invented a fake brand, planted three conflicting stories about it on a blog, on Reddit and on Medium, then asked eight AI tools about it. Some of the tools repeated the invented details, and the brand's own FAQ was one of the sources that kept answers accurate. "When AI has to choose between a vague truth and a specific lie, it often picks the specific lie, unfortunately." Fill every gap about your brand with specific facts on your own site.

Make sure AI crawlers can reach your site

None of the content or mention work matters if AI crawlers cannot reach and read your site. I run four technical checks before anything else.

Check robots.txt for the right bot. This is the setting most guides get backwards. GPTBot is OpenAI's training crawler. ChatGPT search uses a separate crawler called OAI-SearchBot, and OpenAI lets you block one while allowing the other (OpenAI crawler documentation). If you want to appear in ChatGPT search, OAI-SearchBot must be allowed. Google works the same way. Google-Extended covers AI uses outside of search, so Google-Extended does not decide whether you appear in Google Search or in AI Overviews (Google crawler documentation). If your site runs on Cloudflare, check your Cloudflare settings too, because Cloudflare has an option that can add AI bot blocks to your robots.txt file.

Check Google's generative AI setting. In late August 2026, Google added a Search Console setting that controls whether your site appears in Google's generative AI features. Confirm it is set the way you want.

Check whether your text needs JavaScript. Some AI crawlers do not run JavaScript. If your words only load after a script runs, those crawlers may see an empty page. Turn off JavaScript in your browser and reload the page. If the text disappears, you need server-side rendering. Ask your developer, or do what I do and have Claude Code help you make the change.

Check page speed. No AI platform has published an official crawler timeout, but slow pages frustrate people and Google measures that. Aim for Google's "good" threshold, where the main content loads in 2.5 seconds or less (web.dev, Largest Contentful Paint).

You may have heard a lot about llms.txt files. I would skip them for now. If you have time or tokens for one technical fix, spend them on your robots.txt file.

How to track AI visibility

Track AI visibility by measuring citations and lead quality, because AI referral traffic will stay small. AI platforms send far fewer visits than Google does. The quality of those visits is what matters, and that is a reason to track them.

The bigger point is that buyers now form an opinion of you before they reach your site, and some may buy inside ChatGPT. You can have the best welcome offer and the best email flows. If ChatGPT has a poor opinion of you based on what the internet says, those assets never get a chance.

I set up four kinds of tracking:

Track AI referral traffic separately. In GA4, create a custom channel for AI sources such as chatgpt.com and perplexity.ai. Expect an undercount, because some AI apps strip referral data and those visits show up as direct.
Use the free platform reports. Google rolled out a generative AI performance report in Search Console on August 31, 2026. Bing Webmaster Tools has had an AI performance report since February 2026, and it shows how often you get cited.
Read AI bots in your server logs carefully. A bot visiting a page proves the bot can reach the page. It does not prove a citation. To confirm a citation, check the AI answer itself or one of the reports above.
Ask customers. Add "How did you hear about us?" to your forms or checkout and include ChatGPT and other AI assistants as options. Customers who credit AI on that form would often have looked like direct traffic otherwise.

My own tracker taught me one more rule. The first full run reported success. When I opened the data, most of the rows were errors, because I had hit the daily spending cap on the data service the tracker uses. I now check the error rate before I trust a green check mark from any automation. I share the tracker and an SEO and AEO context file for free at jakebauman.io/ai-search-autopilot.

Where I would start

If I were starting AEO from scratch, I would follow the Jake Bauman AEO start order. It takes three steps:

Fix your crawler settings before you write a single new post.
Update one older page that already has backlinks.
Check the Google Search Console and Bing Webmaster Tools AI reports to see what changed.

Once the technical work is done, step back and look at your brand the way a buyer does. Skip your own review widgets, since many of those reviews were prompted or incentivized. Look at Trustpilot, at social media, at forums and at what the press says about you. Then ask ChatGPT directly what people are saying about your brand, whether it can be trusted and whether the product is good.

What bad customer reviews taught me about AI search

Customer sentiment is now an AI search input, and a brand that treats customers badly will get cited badly. I learned this at a large, decades-old franchise company I used to work for. The company had started as a small family business and grown nationwide. When I joined, a private equity firm held a large stake and was nearing the point where it needed to sell.

Margins became the priority. The executive team removed warranties and refunds on an expensive product. I pushed back hard and told leadership the decision would crush the company. Negative reviews came in like a flash flood, and the Better Business Bureau got involved. I have been gone a long time, but a former colleague told me the company is now in an AI search crisis.

You can optimize your SEO and AEO as hard as you like. If you forget the customer, you build bad brand sentiment, and "AI is getting smart and it can develop a sentiment around you just like a human does."

I run NorthSignal, where I work on search visibility and AI citations for owner-led businesses.

This week, ask ChatGPT what people are saying about your brand and whether it can be trusted, and read the answer the way a new customer would.