SEO, AEO, and GEO experts disagree constantly, and most of them aren’t wrong. Google, ChatGPT, Perplexity, and Claude are four separate systems with four different ideas of what makes a source trustworthy, and a study of over a hundred thousand AI-generated answers found only 11 percent of cited sources overlapped across platforms. That’s why one marketer’s tactic can succeed and another’s identical test can fail: they were never measuring the same system. What survives every version of this argument, across a decade of Google updates and now a handful of competing AI engines, is the same thing it always was: verified, consistently trustworthy information wins, regardless of which system is asking.
Everyone's Right, and That's the Problem
Scroll LinkedIn or Facebook for five minutes and you’ll find the same argument, wearing a different outfit each time.
Someone posts that SEO is dead. Forty comments later, someone else posts that SEO isn’t going anywhere, it just looks different now. A third person jumps in to say they’re both missing it: it’s not SEO anymore, it’s AEO. Someone corrects them: no, it’s GEO. Someone else says forget acronyms, it’s entity optimization. Someone says backlinks still win. Someone else says backlinks have been dead for two years and anyone still building them is wasting a client’s money.
It’s exhausting. And if you run a business, not a marketing agency, just a business trying to get found, it’s worse than exhausting. It’s paralyzing. You don’t know who to believe, so you either freeze, or you pick whichever voice was loudest last week and hope.
Here’s what nobody in that comment section seems willing to say out loud:
Almost none of these people are lying.
They’re standing in different rooms of the same house, describing what they see out their own window.
One of them runs an agency working only with roofers. Another manages enterprise SaaS accounts. Another has spent the last six months watching AI Overview citations for a healthcare client. Another has never once checked whether their clients show up in Perplexity, because none of their customers use it. Another swears schema markup changed everything, because it did, for the one platform they were measuring. Another swears schema did nothing, because they were measuring a different platform entirely, and got a different, equally true answer.
They’re not describing the same thing. They just think they are.
Search didn’t die. Google didn’t become irrelevant overnight. AI didn’t replace everything in a single update cycle. What actually happened is quieter and, honestly, more disruptive: the battlefield got bigger, and it split into pieces that don’t all play by the same rules.
Today, a business’s visibility isn’t decided by one system. It’s being decided, at the same time and independently, by traditional Google search, AI Overviews, ChatGPT, Claude, Gemini, Perplexity, and whatever launches next quarter. Five or six different judges, each scoring a different exam, each occasionally disagreeing with the others, and all of them mattering.
So when someone tells you “this worked,” you should probably believe them. Something did work, for their client, in their vertical, on the platform they happened to be watching.
What you shouldn’t believe is the next sentence, the one where they tell you it’s the only thing that works, and that everyone disagreeing with them is a fraud.
That’s the actual argument worth having in this industry. Not who’s right and who’s wrong. But why smart, experienced marketers keep arriving at completely different conclusions from what looks, on the surface, like the same evidence.
Once you understand why that happens, the noise gets a lot quieter, and a lot easier to ignore.
The Fishermen Were Never Fishing the Same Water
How Do Google, ChatGPT, Perplexity, and Claude Actually Differ?
Picture four fishermen. One fishes lakes. One fishes rivers. One fishes far offshore. One fishes small ponds behind people’s houses.
Ask all four of them the same question: “What’s the best bait?”
You already know what happens next. They argue. Not because one of them is lying, and not because three of them are wrong and one is right. They’re catching different fish, in different water, with different conditions, and every one of them is describing something true about the water they actually fish.
That’s not a metaphor stretched to fit. It’s close to a literal description of what’s happening in marketing right now.
For most of the last twenty years, “search” meant one thing: Google. One index, one set of ranking signals, one system to study and argue about. Practitioners could disagree about tactics, but at least they were disagreeing about the same exam.
That’s no longer true. Search today is at least five separate systems, and they don’t retrieve information the same way, don’t trust the same signals, and often don’t even pull from the same pool of sources.
Here is what that actually looks like when you line the systems up side by side.
Google AI Overviews / AI Mode
What It Actually Retrieves From
What It Rewards Most
Average Citations Per Answer
Around 8 per answer
Notable Bias
ChatGPT
What It Actually Retrieves From
What It Rewards Most
Average Citations Per Answer
Notable Bias
Perplexity
What It Actually Retrieves From
What It Rewards Most
Average Citations Per Answer
Notable Bias
Claude
What It Actually Retrieves From
What It Rewards Most
Average Citations Per Answer
Notable Bias
Gemini
What It Actually Retrieves From
What It Rewards Most
Average Citations Per Answer
Moderate
Notable Bias
SYSTEM
What It Actually Retrieves From
What It Rewards Most
Average Citations Per Answer
Notable Bias
Google AI Overviews / AI Mode
Google’s existing web index, powered by Gemini 3
Pages already ranking well organically, semantic completeness, self-contained passages
Around 8 per answer
Increasingly detached from top 10 organic rankings; category answers over brand answers<
ChatGPT
Bing index, only when browse mode triggers
Recency, structured “answer first” formatting, editorial credibility
3 to 6 typical, often none if browse mode does not fire
Perplexity
Its own continuously updated index of 200+ billion URLs
Claude
Gemini
These figures come from a handful of third-party studies published in the first half of 2026, not from official documentation released by each company. Treat the numbers as directionally useful rather than exact, a caveat that matters more than it sounds like it should, since the exact same caution is what this whole article is arguing for.
Google’s AI Overviews still lean heavily on the existing organic index. The system is grounded in the same core ranking and quality signals that have always driven regular search results, and Google has said plainly that there are no special extra requirements to appear in an AI Overview beyond what already makes a page rank well. But even that relationship is loosening. Citation studies have shown Overview citations that once mirrored the top ten organic results detaching from that pattern over the last year, sometimes by a wide margin.
ChatGPT works differently. A large share of its answers never touch the live web at all. When it does search, most of what it retrieves never survives into the final answer, and what does survive skews toward a distinct set of sources: Wikipedia, Reddit, Amazon, and major publishers, refreshed far more recently than what Google tends to surface for the same question.
Perplexity was built from the ground up as a citation engine, not a ranking engine. It searches an enormous, continuously updated index and tends to cite several sources per answer, more than any other platform, with a noticeable lean toward community sources like Reddit.
Claude is the biggest outlier of the group. Rather than crawling and ranking the open web the way a traditional search engine does, it grounds many of its answers in whatever it’s given directly, and when it does search live, it uses a different backend entirely, with a more conservative bar for what it’s willing to cite.
Four systems. Four genuinely different sets of rules. And that’s before you add Gemini, Copilot, and whatever launches next.
Here’s the number that should end most of these arguments before they start: one large study analyzing well over a hundred thousand AI-generated answers across four major platforms found that only about one in nine cited domains showed up on more than one of them. Eleven percent overlap. That means, on average, nine out of every ten sources one engine trusts enough to cite are sources a different engine never touches at all.
Read that again, because it’s the whole point of this article. It’s not that one engine is right and the others are behind. It’s that they are, quite literally, reading different water.
So when a marketer tells you their approach worked, and someone else says the opposite happened for them, there’s a good chance neither of them is wrong. They tested different bait, in different water, and both of them caught fish. The mistake isn’t in what either of them found. It’s in assuming the water they were fishing looks anything like the water everyone else is standing in.
None of this is new, either. The systems are new. The disagreement underneath them isn’t. Marketers have been having a version of this exact argument, in a slightly different costume, for as long as Google has been making changes to how it decides what to trust.
How We Got Here: A Short History of the Same Argument, Repeating
If this argument feels familiar, that’s because it is. The names change. The panic doesn’t.
In 2011, Google released an update called Panda, targeting thin, low-quality, duplicate content and content farms. It hit roughly one in ten search results in the US. People who had built businesses around cheap, high-volume content lost most of their traffic within days. The prediction at the time was the same one you hear now: SEO as we know it is over.
It wasn’t over. It changed.
In 2012, Penguin arrived and did the same thing to manipulative link building, devaluing spammy backlink schemes overnight. Businesses that had built their entire strategy around buying links watched years of work disappear in a weekend. Again: SEO is dead, people said. Again: it wasn’t. It changed.
In 2013, Hummingbird rebuilt how Google interpreted conversational and semantic search queries, moving away from matching exact keywords toward actually understanding what a person meant. In 2015, RankBrain added a machine learning layer to handle ambiguous queries. In 2022, the Helpful Content update introduced a sitewide signal rewarding content written for people first, penalizing content written primarily to game the algorithm.
Then in May 2024, Google launched AI Overviews, generative summaries built on its Gemini models, placed directly at the top of search results. And the same argument started all over again, louder this time, with a longer list of new acronyms attached to it.
Here’s the pattern worth noticing. Every one of these updates was framed, in the moment, as the end of search as we know it. Every one of them was actually a correction toward the same underlying value: reward what’s genuinely useful, penalize what’s trying to trick the system. The businesses that survived Panda, survived Penguin, and are still visible today weren’t the ones chasing whatever loophole the algorithm hadn’t closed yet. They were the ones that happened to already be doing the boring, real thing the update was designed to reward.
AI Overviews, ChatGPT, Perplexity, and Claude are not a break from that pattern. They’re the same correction, applied by more systems at once, moving faster than any single update used to. The businesses still visible five years from now will very likely look like the businesses that survived the last five algorithm shifts: not the cleverest, the most consistently real.
The Two Most Dangerous Words in Marketing... "Studies Show"
Does Schema Markup Actually Help AI Search Rankings?
Two words that end a conversation and rarely deserve to. Testing matters, data matters, but between “something changed” and “here’s why it changed” sits every other explanation nobody bothered to rule out. A page loses word count and traffic climbs the next month. Long content is dead, the post declares. Maybe. Or an algorithm shifted, a competitor lost backlinks, internal linking improved, or the page finally got crawled correctly for the first time in months. Isolating one cause from that many moving parts is usually closer to wishful thinking than analysis.
Nowhere does this play out more clearly, right now, in real time, than the ongoing argument over schema markup, and it’s worth walking through in full because it is close to a perfect case study of everything this article has been arguing.
Here is the stat that launched a thousand agency sales decks in 2026: pages cited by AI search were found to be almost three times more likely to carry schema markup than pages that weren’t cited, based on an analysis of six million URLs. Agencies took that number, built entire service packages around it, and started selling schema markup as the missing key to AI visibility.
Then someone actually tested it properly. A controlled study tracked 1,885 pages that added JSON-LD schema over a seven-month window, matched against 4,000 similar pages that didn’t add it, and measured what happened to citations over the following month, specifically on Google AI Overviews, AI Mode, and ChatGPT.
The result: no meaningful increase. The AI Mode and ChatGPT numbers landed close enough to zero that researchers classified them as statistical noise. The correlation from the earlier six-million-URL study was real. The causation wasn’t. Pages that carry schema tend to be built by more technically capable teams, publishing stronger content, earning more links, and already ranking better in regular search. The schema wasn’t creating the visibility. It was simply riding along next to it, on the same pages, for unrelated reasons.
And even that isn’t the complete picture, because a separate large-scale rollout across more than 2,000 URLs, presented at a major SEO conference in early 2026, found the opposite result on a different platform. After adding schema sitewide, Google AI Overview citations increased by over 1,500 percent. AI Mode citations increased by 377 percent. In the exact same rollout, on the exact same pages, citations on ChatGPT and Gemini dropped, and Perplexity showed no change at all.
Same tactic. Same pages. Four completely different outcomes, because it was never one test being run. It was four separate systems being asked the same question, and only one of them happened to respond to that specific piece of bait, in that specific water.
This is precisely why two honest agencies can run two honest tests and walk away with two opposite, equally true conclusions. One was measuring Google AI Overviews, where schema genuinely moved the needle. The other was measuring ChatGPT, where it didn’t move at all. Neither one is wrong. Neither one has actually proven anything universal. Each of them proved something real about the one system they happened to be standing in front of.
The most accurate synthesis of the schema debate, from a source that actually reviewed the competing evidence rather than picking a side, put it plainly: schema could plausibly act at several different points in the retrieval pipeline, so a flat claim that schema does nothing is too strong, in the same way a flat claim that schema is the answer is too strong. The honest position sits in the middle, and the honest position is rarely the one that gets shared.
This Stopped Being About Search a Long Time Ago
Watching that same argument repeat for over a decade, I stopped believing the fight was really about algorithms.
I think it’s about identity.
If you’ve spent fifteen years building a reputation on backlinks, what happens the moment someone says backlinks don’t matter anymore? If your whole agency is built around technical SEO audits, how do you react when someone claims AI changed everything overnight? If you’ve spent years teaching topical authority frameworks, how does it feel to watch an AI cite a page that ignores your entire model and still wins?
It feels personal. Because at that point, it isn’t really the strategy being questioned. It’s the career built on top of it. The expertise. The credibility. The years spent becoming the person people call when they have this exact question.
Social media doesn’t help. It rewards the opposite of nuance. Nobody shares “it depends.” Nobody screenshots “in our testing, results varied by industry.” Those posts get scrolled past in half a second. What gets shared is “everything you’ve been doing is wrong,” because certainty gets clicks and nuance doesn’t, and the platform people are arguing on has its own separate incentive completely unrelated to the actual question of what works in search.
That’s worth sitting with for a second. The people yelling the loudest about which search algorithm rewards which behavior are, in that exact moment, being rewarded by a completely different algorithm, one that has nothing to do with SEO and everything to do with engagement. The confidence isn’t free. It’s being paid for, just not by the platform they’re claiming expertise on.
Meanwhile, the marketers I trust most tend to talk the least like that. They say “maybe.” They say “it depends.” They say “in our testing” instead of “studies show.” They say “we’re still learning,” even after years in the business.
That’s not hedging. That’s what actual experience sounds like once someone has been burned enough times by a result that didn’t repeat the way they expected it to.
The industry doesn’t reward that kind of honesty with attention. But it’s usually the honest ones who are still right a year later, long after the confident hot take has quietly been deleted.
Becoming the Answer
What Is GEO, and How Is It Different From SEO?
Here’s the shift worth paying attention to. The question used to be “how do I rank?” That question is starting to matter less than a different one: “how do I become the answer?”
Those sound similar. They’re not. Ranking is about winning a position on a page. Becoming the answer is about being trusted enough, by enough different systems, that you get named directly, sometimes without a click at all.
Look at what each engine is actually optimizing for underneath its own retrieval logic. Google wants results people trust. ChatGPT wants sources people trust. Perplexity wants citations people trust. Claude wants context it can verify. Gemini wants information it can stand behind. Different mechanisms, different retrieval paths, same destination every time: trust.
That pattern shows up in the data too. One large study analyzing over seventeen million AI citations across four major engines found that, despite each one retrieving information in a completely different way, verified business listings and directory presence accounted for more than half of all distinct citation sources. Not website copy. Not markup. Verified, claimed, accurate records of a real business, sitting in the places every engine ends up checking anyway.
Worth being honest about that number. It comes from a single vendor whose own product is listings management, so treat it as directionally useful rather than settled proof, the same caution worth applying to almost any single study in this space, schema included. But directionally, it lines up with everything else in this article. The systems disagree on almost everything. The one thing they keep agreeing on is that a business that’s verifiably, consistently, boringly real tends to win more often than one chasing a clever trick.
That’s the part that should be good news, not bad news, for the kind of business that’s been doing things the right way and wondering why it doesn’t feel like it’s paying off yet.
I’ve watched this play out with a client who was, for practical purposes, invisible everywhere that mattered: inconsistent information, no real digital footprint, nothing an algorithm or a customer could actually verify. The fix wasn’t a trick. It was rebuilding the boring stuff: consistent, verified information across every place a customer or a system might look, paired with content that actually answered the questions people were asking. That business went from invisible to impossible to miss, and that growth funded expansion into new locations across multiple states.
That’s the actual work now. Not chasing whichever acronym is trending this month. Building something that every search engine, every AI model, and every real customer would look at and say the same thing: this is who I’d recommend.
So instead of asking what the next SEO trick is, ask a better question. The question the smartest marketers in the room have been asking for a long time…
What would make every search engine, every AI model, and every customer trust us enough to recommend us?
That question doesn’t go out of date when the next algorithm update ships. It’s the only one that hasn’t changed in twenty years, and it’s the only one that’s going to matter twenty years from now.
So… What Do We Actually Do?
Chas and the team have spent years building C.H. Local Media, a marketing agency working with local service businesses, HVAC companies, insurance agencies, moving companies, tree services, towing operations, and financial planners, on the exact problem this article describes. Getting found across a search landscape that no longer plays by one set of rules.
He’s watched a single visibility rebuild take clients from effectively invisible to operating in multiple states, not through tricks, but through the same boring, verified fundamentals this article argues actually work. Long before AI search existed, he learned a version of this lesson the hard way, as a drummer in a regional touring band, where credibility wasn’t claimed, it was built night after night in front of people who’d never heard of you yet.
That’s the lens this article comes from. Not theory, and not someone selling the acronym of the month, but someone who has to make this stuff work for real businesses, on real budgets, every day.
