Start AEO with Perplexity: our engine data says it cites you first

Checking where an AI engine cites your own pages is the first baseline a small business can take.

In one of our recent pilots, we read a software product end to end, from the admin setting down to the screen a user actually sees, and found the two quietly disagreeing: the admin page described two-factor sign-in as using an authenticator app, while the real sign-in screen sends a one-time code by text and email. I have thought of that small discrepancy often since. What we believe our pages say, and what a machine finds when it reads them, are not always the same thing.

That gap is where AI search begins for a small business. Which of your pages would an AI engine cite today, if a customer asked the question you hear most often? My argument begins there; the research section below walks through the evidence, and our look at how each AI engine cites the web fills in the wider picture.

The mechanics are not mysterious. Working marketers already repeat them: put the actual answer in the first sentence, keep paragraphs short, add FAQs and schema, cut the filler. What they rarely say is which engine will notice first. That question decides where a small site spends its first hours, and the answer lies in how each engine chooses the pages it reads.

Why does the choice of a first AI engine matter for a small business now?

The first engine you measure decides how soon a small site sees proof of progress. We built our view of that order from 26,577 real AI-visibility audits.

Before a single hour is spent on pages or prompts, a small business owner can settle three things in plain words:

1. Write down the questions customers actually ask before they buy, in their own phrasing.
2. Choose one engine to measure first, and measure it the same way every week.
3. Decide in advance that a first citation of your own page counts as a real signal, not a curiosity.

Search has shifted the way an underground river shifts, quietly and beneath notice, until one morning the old channel is dry and the water is running somewhere else. In 2025, Jonathan Mast, who began in digital marketing in 1995, told the Public Relations Stories and Strategies podcast that he does not "remember the last time I actually used Google to do a search." His co-host, Farzana Baduel, said she had stopped using Google as well and now asks ChatGPT "everything." These are marketers, not casual users. Their habits tend to arrive early.

The common assumption is that a small business should begin where the crowds are, and so it begins with ChatGPT. I think that confuses the size of a sea with the nearness of its shore. A crowd is not a citation. What a small site needs first is the engine most likely to notice the pages it already has, because an early, visible signal is what keeps an owner investing through the long quiet months when nothing seems to move.

Our team has scored 11,000+ domains across 15 sectors and 28 categories, and that work runs on infrastructure designed by a full-stack engineer and content infrastructure architect with 20 years of building enterprise systems. Laid side by side, the 4 pieces of evidence in this section describe one drift: buyers now put their questions to machines, and someone has to watch where the answers come from. The shape of those questions is changing too. On Lenny's Podcast in 2025, Ethan Smith passed along a loosely sourced figure, which he thought came from Perplexity, that the average chat prompt ran around 25 words, against around six words for a Google query. A customer who once typed a few nouns now describes a whole situation, with its place and its urgency folded in. The number of people asking this way was growing as well. In 2025, Jim Kunkle reported on The Digital Revolution that 13 million U.S. internet users had relied on generative AI as their primary search tool in 2024, a number projected to exceed 90 million by 2027. For a small business, that means the first engine worth studying is one whose answers can be traced back to a page, so you can see which of those long questions already lead to you. That is the purpose of research-driven content built on evidence, which treats each claim as something an engine can trace.

The stakes are simple. An answer your customer reads without clicking still shapes the sale. A business absent from that answer is absent from the decision.

So the question is not whether to begin, but where the first light is likely to fall, and the engines, it turns out, do not choose their sources in the same way at all.

If buyers ask ChatGPT first, why not start optimizing there?

Ask which engine lets you see, page by page, where its answers come from. We commit to getting clients named in AI answers in 100 days, so the order of engines matters.

Here is the friction, and I do not want to pretend it away. The people a small business hopes to reach are, in large numbers, opening ChatGPT before anything else. Jenna Nelson, founder of HerAgency, said as much on the Disambiguation podcast in March 2026: "I go to ChatGPT first, or maybe perplexity." On the same episode, the host, Michael Fauscette, described Perplexity as the default search engine in his browser. Two experienced marketers, one conversation, two different first doors.

That split is the whole tension in miniature. Demand gathers around the largest assistant. Citations, the moment your own page is named as a source, may gather somewhere else first. A small business with a few hours a week cannot chase both at once, and so it must decide which signal it will trust while the other one is still dim.

Nelson's own advice leans toward breadth. She argues that businesses should combine SEO, AEO and GEO "so that you're not being missed anywhere that your audience might be looking," and she believes small businesses and solo founders can make a big splash now because larger competitors are slow to move. I agree with the breadth as a destination. I part ways with it as a starting line.

Breadth is expensive. A first signal is cheap.

Alex and I bring a combined 40+ years of SEO and content infrastructure experience to AEO, and if that long apprenticeship has taught us anything, it is that the work which survives is the work someone could measure early enough to keep funding. The outcome we point to is modest in scale and plain in shape: one client went from a quiet site to 10 booked calls in 3 weeks. Quiet sites do not wake all at once. They wake in one place, and then another, the way lights come on across a dark valley one farmhouse at a time.

So the practical stance is not ChatGPT or Perplexity forever. Ethan Smith advised in 2025 choosing engines by "which surfaces have the most traffic and then track those," a fair case for the crowd, and one I take seriously. The guides in our AEO knowledge base of AI search guides treat each engine separately for exactly this reason. Owners who skip the first step tend to read silence in ChatGPT as failure, when it may only mean the light has not reached that window yet, and what decides where it lands first is something older than any of these engines.

Which engine should a small business trust as its first signal?

Trust Perplexity as the first gauge. Read which of your pages it cites for the questions customers really ask; the research section below walks through why it comes first.

I founded LiveHelpNow, an Inc. 5000 inductee (#84, 2015-2018), long before anyone spoke of answer engines, and the oldest lesson of that work still holds in this new country: put your first effort where the ground already bears weight. For AI search, that ground is the organic standing a business has built, slowly, page by page, over years.

A study by Graphite of thousands of questions, described by Ethan Smith in September 2025, found that citation overlap with Google search results was around 70% for Perplexity and around 35% for ChatGPT. That gap is the hinge of this whole argument. Before leaning on that figure, test it against your own site: search Google for one customer question, note the pages that rank, and see whether Perplexity cites the same ones. An engine that draws less on those results asks for something wider before it names you.

Jeff Slater, VP of Operations at the orthodontic marketing firm Kaleidoscope, put the lineage plainly on the Orthodontic Products Podcast in 2025, calling AEO and GEO "subsets of what we think of as traditional SEO." His team reverse-engineers AI Overviews by running local searches, finding the competitors already cited, and studying how those cited passages are formatted and placed. Ethan Smith laid out a similar sequence on Lenny's Podcast in 2025, and it translates well to a business with a modest budget. Take the phrases you would pay for in search ads, the ones that bring in paying work, and turn each into the question a customer would actually ask. Track those questions over time. Note who shows up as the cited source for each one, build your page in whatever form already dominates those citations, and then answer every follow-up question a buyer is likely to ask next. Smith also offered newcomers a rare consolation: "you can get mentioned by a citation tomorrow and start showing up immediately," through a Reddit thread, a YouTube video or a mention on someone else's blog. Each of those mentions is a page an engine can cite, and a page a small site can actually check. Small practices can copy that habit tomorrow.

The conclusion follows. Your existing pages are not wasted. They are the bones on which a first citation grows.

Our AEO Rank standard scores every site on 53 criteria across 5 pillars, and we track whether ChatGPT, Claude, Perplexity and Google AI Overviews name our clients. What our records do not yet hold, in a form I can publish, is a count of the queries where Perplexity cites a client site that ChatGPT skips, or the elapsed time between a first citation in one engine and a first in the other. Those are the right measurements. I would rather leave the space empty than fill it with a guess. A fuller set of answers on AI search optimization covers how we score each pillar.

For software companies, the same principle reaches down into the code itself. We connect the repositories a product is built from, re-index them every night, and propose knowledge-base questions users actually ask, each answered from the code alone. A page written from the true source, rather than from memory, is the kind of page a citation engine finds and keeps, and the first place to watch that happen is a weekly check of a handful of real questions in Perplexity.

What should a small business do first, starting this week?

List the five questions clients ask you most often, put each to Perplexity this week, and keep a dated note of which page, if any, it names.

In 2025, Jim Kunkle's podcast The Digital Revolution cited a Gartner projection that 30% of online searches would be AI-driven by 2026, and that year is nearly spent. The same 2025 account put ChatGPT Search at 4 billion monthly visits. I will not pretend that crowd is unimportant. The research section above weighs that crowd against the signal. It is simply not the nearest shore, and a small business that sails straight for the far coast often runs out of patience before it lands.

If no page answers a question in two or three plain sentences near the top, write that answer first. Working marketers agree on the craft: put the answer first. What I would add is an order of travel.

Begin with one question, one engine, one page that answers it plainly, and let the next citation tell you where to go.