A practical guide to earning citations from ChatGPT, Google AI Mode, and Perplexity, so the question of how to get AI to recommend your business stops being a mystery and starts being a process you run.
Ranking Made You Findable. Getting Quoted Makes You The Answer.
For twenty years the game was simple. You published something useful, you climbed the results page, and people clicked. Position one was the prize because position one got the traffic. That whole model rested on one assumption: that the person searching would scroll down and click a link.
That assumption is quietly dying, and most entrepreneurs have not adjusted.
Here is the direct answer to the headline. You get AI to cite your business by publishing the specific things a model cannot invent on its own: your original numbers, your named clients, your documented process, your dated results, your direct quotes. Then you structure those things so a machine can lift them cleanly, and you make sure they also exist on sources outside your own website. Competent generic content is now infinitely available and worth close to nothing. Specificity is the only currency an AI cannot counterfeit.
That is the whole shift in one sentence. The win condition is no longer being found. It is being named.
This is not a doom post. AI search is not the end of content marketing, it is a re-pricing of it. The thin, competent, keyword-shaped article that used to earn a middling ranking is now worth almost nothing, both to a reader who has seen a thousand of them and to a model deciding whom to cite. Meanwhile, the piece with one number nobody else has and one client story nobody else can tell just became far more valuable than it was two years ago.
Most people will respond to this shift by publishing more. That is the wrong lever. The right lever is publishing what only you can prove.
Key Takeaways
- Search behavior has changed at the click level, and Pew Research Center data shows users click a traditional result far less often when an AI summary appears above it.
- The new win condition is citation, which means being named inside the answer rather than sitting in a list of links beneath it.
- Academic research on generative engine optimization found that adding statistics, direct quotations, and cited sources measurably increased how often content got pulled into AI answers.
- Pangram Labs found that more than 40 percent of longform LinkedIn posts in its sample were fully AI-generated, which means generic competent content is now abundant and cheap.
- The content that earns citations is the content a model cannot synthesize: your first-party data, your named clients, your documented method, and your dated results.
You Optimized for a Click that Fewer People are Making
Think about how you personally used a search engine last week.
You typed a question. An AI-written answer appeared at the top. You read it. And then, if you are like most people, you closed the tab. You did not scroll. You did not click through to the seventh-best blog post on the topic to verify. The answer was right there, it was good enough, and you moved on with your day.
Now flip that around and look at it from the business side. Every piece of content you published in the last five years was built for the other behavior. It was built to earn a position, catch an eye, and get a click. Everything downstream, the email capture, the offer, the sale, assumed a human being would land on your page.
When the click stops, the chain breaks. And here is the part that stings: your content can still be excellent, still rank well, and still lose. You can hold position one and watch traffic decline anyway, because a summary above you already answered the question using your ideas without sending anyone your way. It is not that your content got worse. It is that the payoff moved.
There is a second problem stacked on top of the first, and it is the one almost nobody is planning for. The supply of decent content has exploded. Anyone with a subscription and ten minutes can now produce a competent, well-organized, perfectly readable post on any topic in your industry. Pangram Labs, which builds AI detection tools, analyzed more than a million social posts and found that AI writing is now everywhere in the feed, with longform content hit hardest.
So the market is flooded with content that is fine. Not bad. Fine. And fine has been commoditized to zero.
Combine the two problems and you get the real picture. Fewer people are clicking, and the content that used to earn clicks can now be produced by anyone in minutes. If your plan is to publish more competent articles on topics everyone else is also covering competently, you are competing where supply is infinite and value is zero.
But what if the same shift that devalued generic content made your specific, hard-won, first-party knowledge more valuable than it has ever been?
What the Data Actually Shows About Clicks and Citations
Let me walk you through the numbers, because the story they tell together is sharper than any of them alone.
Start with the click. The Pew Research Center analyzed the browsing behavior of 900 U.S. adults across 68,879 Google searches conducted in March 2025. When an AI summary appeared in the results, users clicked a traditional search result link in 8 percent of visits. When no AI summary appeared, they clicked in 15 percent of visits. Roughly half the clicks, gone. Pew also found that only 1 percent of visits to a page with an AI summary produced a click on a link inside that summary, and that browsing sessions ended after the search 26 percent of the time when an AI summary was present, compared with 16 percent when it was not.
Ahrefs looked at the same phenomenon from the keyword side. Analyzing 300,000 keywords, Ahrefs reported in April 2025 that the presence of an AI Overview correlated with a 34.5 percent lower average click-through rate for the top-ranking page compared with similar informational keywords with no AI Overview. Position one, same page, fewer clicks.
Now the supply side. Pangram Labs published research on July 9, 2026, based on 1,002,627 posts scanned across LinkedIn, Medium, Substack, X, and Reddit through its browser extension. The average AI rate across all scanned items was 13.8 percent, but the concentration in longform was much higher: about one in four items over 250 words flagged as fully AI-generated, and on LinkedIn more than 40 percent of longform posts flagged as fully machine-written. Substack integrated Pangram’s detection directly into its platform in July 2026, according to Axios reporting, so that readers can see whether a post was human-written.
Read those two sets of numbers together. Attention at the link level is contracting, and the supply of competent generic writing is expanding. That is a brutal combination for anyone whose content strategy is volume.
Here is the number that points to the way out. The original academic work on generative engine optimization, published by researchers from Princeton, IIT Delhi, Georgia Tech, and the Allen Institute for AI at KDD 2024, tested nine content tactics across roughly 10,000 queries. The tactics that most reliably increased visibility in AI-generated answers were adding statistics, adding direct quotations, and citing sources, each producing improvements in the range of roughly 30 to 40 percent. The effect was largest for sites that were not already dominant, with the cite-sources method producing a large visibility gain for a site sitting fifth in traditional search results.
In other words, the machine rewards provenance. It pulls in the sentence that carries a number, a name, and a source, because that sentence is the one it can safely repeat.
One more data point on why this is worth the effort even with fewer clicks. A Semrush study reported in June 2025 estimated that AI search visitors were roughly 4.4 times more valuable than traditional organic search visitors, on the logic that someone arriving after an AI conversation has already done their comparison shopping. Fewer visitors. Warmer visitors.
And Andy Crestodina of Orbit Media, who scanned 150 SEO articles about AI search optimization for his April 2026 analysis, found that 93 percent recommended FAQs and 89 percent recommended schema markup. Everyone is running the same checklist, which means the checklist is not the advantage.
Publish the Things A Model Cannot Make Up
The mental shift is this. Stop asking “how do I rank for this topic?” and start asking “what would a model have to quote me to say?”
A large language model generating an answer is doing something closer to writing a research summary than ranking a list. It needs facts it can stand behind. It reaches for claims that are concrete, attributable, and specific, because those are the ones that survive the process of being compressed into an answer. A paragraph of well-written generalities gives it nothing to hold onto. A sentence like “in our review of 340 onboarding calls in the first half of 2026, 61 percent of clients named the same single bottleneck” gives it something it cannot get anywhere else.
That is the asset. Not your opinion about the industry. Your evidence about it.
You already own more of this than you think. Your sales call transcripts. Your support tickets. Your onboarding surveys. Your client results. Your before-and-after numbers. Your pricing experiments. None of it exists anywhere in a training set, which is precisely why it is valuable.
Crestodina makes this point well. He found that 78 percent of the SEO articles he analyzed told readers to source their FAQ questions from SEO tools, while only 4 percent suggested asking their own internal teams. Just three of the 150 suggested the approach he recommends: loading your actual sales call transcripts into an AI tool and extracting the questions your real buyers ask, in their own words. As he puts it, the most effective method is the least likely to be recommended.
Lily Ray, quoted in that same Orbit Media piece, adds the companion point: unique, original research is what naturally earns discussion and third-party citation, while structured data and extractable content blocks should already be table stakes.
Brian Piper, Director of Content Strategy at the University of Rochester and co-author of the second edition of Epic Content Marketing, frames the same transition as a move from SEO to AEO, answer engine optimization, built on long-form, human-first content designed for AI discovery. Notice the phrase “human-first.” The path to machine citation runs through content a human expert actually had to live to produce.
So the solution has two halves, and you need both.
The first half is substance. Publish original evidence. Your numbers, your named clients with permission, your documented process with real steps, your results with dates attached. Specificity a model cannot synthesize.
The second half is structure. Make that substance easy to extract. Answer the question in the first hundred words. Use question-shaped headings. Put your key finding in a single clean sentence that could be lifted whole. Cite your own sources. Add the date. Name the sample size. A model that has to work to find your claim will use someone else’s.
Substance without structure gets ignored. Structure without substance gets skipped. Together, they get you quoted.
Practical Steps to Get Your Business Cited by AI
Here is the process, in order. You can start this week.
1. Find the prompts your buyers actually type. Do not guess. Generate at least ten realistic, commercial-intent prompts your ideal client would type when evaluating options, run each one through ChatGPT, Google AI Mode, and Perplexity, and write down which brands get named and which sources get cited. This is your real competitive landscape now, and it is category-specific.
2. Mine your own conversations for the real questions. Export your sales call transcripts, support tickets, and onboarding notes into a single file and load them into an AI tool. Ask what questions come up most, what prospects get wrong about you, and what words they use before they know the technical term. That is your FAQ, and nobody else has it.
3. Publish one number nobody else can publish. Every quarter, run a small piece of original research from data you already own. A count, a percentage, a before-and-after, a survey of your own clients. One real finding, clearly stated with its sample size and date, outperforms ten opinion posts.
4. Answer the question in the first hundred words. Lead with the direct answer in a clean, self-contained sentence, then prove it for the rest of the piece. If a model has to read eight paragraphs of throat-clearing to find your point, it will quote whoever said it in one line.
5. Name names, dates, and sources. Attach a date to every result. Name the client where you have permission. Cite the study, the tool, the version. The research is clear that quotations, statistics, and citations are what get content pulled into AI answers, so give the machine something citable on every page.
6. Get quoted somewhere that is not your website. Models weigh third-party sources heavily, and in Crestodina’s summary of the prevailing consensus, the commonly cited figure is that around 85 percent of AI brand citations come from domains other than your own. Podcast interviews, industry roundups, review sites, and expert quotes in other people’s articles all count. Pitch three per quarter.
7. Track citations, not just rankings. Once a month, re-run your buyer prompts and record whether you were mentioned, how you were described, and what got cited instead of you. That log is your new analytics dashboard. Fix the gaps it reveals, and refresh your cornerstone pages on a schedule so the evidence stays current.
Frequently Asked Questions
What is the difference between SEO and AEO or GEO?
SEO optimizes for position in a list of links. AEO and GEO optimize for inclusion inside a generated answer. The tactics overlap, but the goal differs: instead of earning a click from a ranked page, you are earning a mention and a citation within the response itself.
Does getting cited matter if nobody clicks through?
Yes, for two reasons. Being named inside an answer is recommendation, not just visibility, and it shapes the buyer’s shortlist before they ever visit a site. A Semrush study reported in June 2025 also estimated AI search visitors are about 4.4 times more valuable than traditional organic visitors.
How do I know whether AI is mentioning my business right now?
Run ten realistic buyer prompts through ChatGPT, Google AI Mode, and Perplexity, then record which brands appear and which sources the answer cites. Repeat monthly and keep a log. That comparison tells you both where you stand and which sources you need to appear on.
Will using AI to write my content hurt my chances of being cited?
The risk is not the tool, it is the sameness. Pangram Labs found more than 40 percent of longform LinkedIn posts in its sample were fully AI-generated, so machine-shaped generic writing is now abundant. Use AI to structure and edit, but supply the original evidence yourself.
How long does it take to start showing up in AI answers?
Longer than a ranking change and shorter than you fear, and it depends heavily on your category and existing footprint. Treat it as a quarterly discipline: publish original evidence, earn third-party mentions, and re-measure your buyer prompts monthly rather than expecting a single post to move it.
Stop Competing Where Supply is Infinite
Go back to that moment I described at the top, the one where you read the AI answer and closed the tab. You did not click because you did not need to. The answer was sufficient.
Now ask the only question that matters for your business: when your buyer has that same moment, whose name is in the answer they read?
That is the whole game now. Not whether you rank. Not whether you published this week. Whether the model, assembling an answer for someone who will never scroll, reaches for your evidence and says your name.
You cannot win that by publishing more of what everyone else can produce in ten minutes. Competent, generic, well-organized content is now free and infinite, and things that are free and infinite do not earn citations. What earns a citation is the stuff you had to actually do the work to know: the number from your own client data, the process you documented after doing it two hundred times, the result with a date and a name attached.
The good news is that this is the one competition where your years in business are an unfair advantage and your competitor’s AI subscription is not. You have the receipts. Most people never publish them.
So publish them. One real number a quarter. One documented process. One named result, structured so a machine can lift it cleanly.
Rankings made you findable. Evidence makes you quotable. Publish what only you can prove.
If you want the exact prompts, research workflows, and AI systems we use to help entrepreneurs get named in AI answers instead of buried beneath them, that is what we teach inside White Beard Strategies. Grab a training replay or join the membership, and start publishing the things a model cannot make up.
Jonathan Mast is the founder of White Beard Strategies, where he helps entrepreneurs build practical AI systems that create leverage instead of busywork. He is a speaker, a builder, and a longtime believer that the fastest way to stand out in a world of infinite content is to publish the one thing nobody else can prove.
