Why Is My Brand Invisible on Google but Cited on Claude?

Updated September 2026. I wrote in April that no one had this figured out and the data was early. Both things were true, and five months later several of the numbers had already been overtaken. I’ve refreshed them below and added what’s changed. The argument hasn’t changed — if anything, it’s stronger.

—

I was researching sauna manufacturers in BC. (I’m on that bandwagon). I ran the same query across Google, Claude, Perplexity, and Gemini to compare outputs.

I got four different answers.

Not minor variations; I got entirely different sets of “top” companies. Different brands, different narratives, different winners.

The company I ended up going with — a small, well-run manufacturer with a genuinely strong product — barely showed up on Google. Claude surfaced them immediately, with clear reasoning: materials, craftsmanship, and build quality. My criteria definitely influenced the results — I explicitly wanted something designed and built in the Pacific Northwest, using BC western red cedar, and local-ish. Claude also remembered previous searches, so that was a factor… but more on all this below.

I’ve been doing marketing audits and digital strategy long enough to know how SEO works. Or worked. But I couldn’t explain, cleanly, why the company that earned my trust through Claude was effectively invisible in traditional search.

That gap is what pushed AI visibility from “interesting” to “important” for me. It comes up in every marketing audit now. And yet most of my clients have no idea how they show up in AI responses — or whether they do at all.

I’ve followed Rand Fishkin for years (big Moz fan here) and, more recently, Kevin Indig. Between their work and everything else I could get my hands on, I understand this a bit more. This post is a synthesis of what I’ve found useful — aimed at marketers and senior teams trying to decide what to do next.

What Are We Actually Talking About?

The terminology is messy. GEO, AEO, LLMO, GSO, pick your acronym. It’s the usual soup of terms designed to sound smart.

Here’s how I’m thinking about it:

GEO (Generative Engine Optimization) — Making your content citable by AI systems. You want to be referenced, recommended, or included when someone asks a relevant question in Claude, ChatGPT, Perplexity, or Gemini.

AEO (Answer Engine Optimization) — A subset of that: structuring content so it can be pulled directly into answers. Think “featured snippets,” but for AI synthesis.

SEO — Still critical. Still foundational. Also insufficient on its own.

The mistake right now is treating these as either identical or completely separate. They’re neither. There’s overlap, but the outputs are different enough that you need to be deliberate.

Why You’re Getting Different Answers Everywhere

You run the same query across platforms and get different answers. The instinct (mine anyway) is to assume one is “right.” They’re not converging because they’re not optimizing for the same signals.

Based on the best available data (largely from Kevin Indig’s research, which analyzes hundreds of thousands of real ChatGPT conversations rather than theorizing), traditional SEO metrics like domain authority, backlinks, and page speed correlate weakly with AI citations. Current studies suggest different citation patterns across the platforms:

•       Perplexity seems to lean toward depth and completeness. Dense, comprehensive content wins.

•       Google AI Overviews may tend to favour breadth of factual coverage. More specific facts, more inclusion.

•       ChatGPT appears to correlate strongly with brand prominence and readability. Clear language matters. So does your overall brand footprint across the web.

•       Claude and Gemini seem to reward structured, well-reasoned, “sourceable” content.

The divergence isn’t theoretical anymore. Ahrefs measured 76% of pages cited by Google’s AI Overviews as also ranking in the top 10 organic results in its earlier study. Its latest analysis puts that overlap at 38%, although the methodology and dataset changed between studies, so it isn’t a perfect apples-to-apples comparison. Even so, the direction is hard to ignore: ranking well and being cited by AI are no longer the same game even if they’re still played on the same field.

The sauna company I bought from had a detailed FAQ, plain-language product descriptions, clear comparisons, and little marketing fluff. Nothing about it was “optimized” in the traditional sense. But it was easy to parse, specific, and credible. Claude could use it. Google largely ignored it.

Big gap there.

One more thing has shifted since I started paying attention: the platforms themselves are fragmenting. A year ago, ChatGPT was roughly three-quarters of all generative-AI website visits. By mid-2026 it was down to about half — not because ChatGPT shrank, but because Gemini climbed past a quarter of the category and Claude grew from a rounding error to nearly 10% [Similarweb, June 2025–May 2026]. ChatGPT still drives the vast majority of AI referral clicks, but usage (meaning where the recommendations actually happen) is spreading out.

When ChatGPT represented roughly three-quarters of generative-AI web usage, measuring it alone gave you a reasonable proxy for the category. It no longer does.

AI Sends Me No Traffic, and My Analytics Look “Fine” – Who Cares?

Your brand can be absent from AI answers and still look “fine” in your analytics. AI platforms send almost no traffic — still under 1% of all website visits, a number that has barely moved since I first wrote this. What has moved is everything around it.

SparkToro’s June study, built on Similarweb clickstream data, found that 68% of Google searches in early 2026 ended without a click — up from about 60% in 2024. More than two-thirds of U.S. Google searches now end without a click. Add the AI platforms answering questions before anyone searches at all, and the picture is clear: you’re being seen more and visited less.

Decisions are being made upstream. By the time someone Googles your brand name, the shortlist is already formed. AI is acting earlier in the funnel: as a filter, not a channel.

Here’s the part I underweighted in April: the trickle of traffic AI does send is good. Semrush pegs the average AI search visitor at roughly 4.4x the value of an organic search visitor, based on conversion rates. That tracks with what I’ve seen — someone arriving from a Claude or ChatGPT recommendation has already been filtered, compared, and pre-sold. Small pipe, high-intent water.

So “AI sends no traffic” is true, but it's also the wrong test. Better metrics to add (I think) are citation frequency, share of mentions, and how accurately your brand is described, and whether those are improving or eroding over time. Traditional analytics won’t show you any of it.

I ran this on a few client sites recently, and the gap between their Google performance and their AI presence was surprising. One ranked well for competitive terms but barely showed up in ChatGPT, Claude, or Perplexity for the exact questions their customers were asking. They had no idea, and neither did I.

Rand Fishkin Knows What’s Up

Fishkin’s data is worth taking seriously because it’s grounded in actual clickstream behaviour, not surveys or posts like this.

A few things worth noting:

•       Google isn’t dying. Per Fishkin’s analysis, search volume grew 22% in 2024 — roughly a trillion net new queries, more growth than the previous seven years combined.

•       AI platform usage is growing faster — but not sending traffic out. The platforms are designed to answer, not refer.

•       SparkToro can now show you which AI platforms your target audience uses and what prompts they’re using — which is genuinely useful for prioritizing where to focus.

Go where your audience pays attention. That is your new job.

For a growing number of categories — especially B2B and considered purchases like, say, a sauna — that now include AI systems. The job hasn’t changed. The surfaces have.

The question is no longer simply ‘How visible are we in AI?’ It’s ‘Which AI systems influence our buyers, for which questions, and which sources do those systems trust?’ 

The Google AI Mode Wildcard 

When I wrote this in April, I treated “AI search” as ChatGPT, Claude, Perplexity, and Gemini, with Google’s AI Overviews bolted on. That framing is already dated, because Google is building something else inside search itself: AI Mode.

The numbers are very big (and a bit weird) and worth looking at. Google says AI Mode passed a billion monthly users within a year of launch, with query volume more than doubling every quarter. Yet SparkToro’s clickstream data found only about 0.3% of Google searches flowed into AI Mode in early 2026. Both things are true: enormous headline adoption, tiny share of real search behaviour — so far. That may simply be what the bottom of an adoption curve looks like.

Google is also wiring commerce directly into it. At I/O in May, they announced Universal Cart and an open agentic-commerce protocol co-developed with Shopify, Walmart, Target, and others — meaning the AI doesn’t just recommend; it can transact. That’s mostly a B2C story today, but the direction of travel matters for everyone: the answer layer is becoming the action layer.

What to do about it: nothing separate. The same qualities that get you cited in Claude — extractable answers, clean structure, entity consistency, real specifics — are what AI Mode draws on. Google itself published guidance this year saying, in effect, that optimizing for generative AI is doing good SEO, not a new discipline. I half agree. The execution overlaps heavily; the measurement doesn’t.

What Works?

You don’t need to rebuild your strategy from scratch. But you do need to adjust how you execute it. I’m still working through what this looks like in practice. I’m not sure anyone has a complete playbook yet, but here’s what the “evidence” consistently points toward.

1. Make your content extractable.

Answer the question at the top. Use clear headings and FAQ sections. Write in plain English. Go deep on topics that matter. If an AI can’t confidently pull an answer from your page, you won’t show up. This sounds basic because it is — the difference is that it used to be optional.

2. Build entity-level authority.

AI answer systems draw on signals and sources well beyond your own site. They build a picture of your brand across the web: LinkedIn, Google Business Profile, reviews on G2 or similar platforms, directory listings, and — more than most teams are comfortable with — Reddit. Indig’s dataset shows that a 10% increase in G2 reviews correlates with a 2% increase in AI citations, and that Reddit accounts for nearly half of all Perplexity citations. Consistency across those surfaces matters more than polish on any single one.

3. Invest in earned media.

AI systems encounter earned media through a mix of training data, search indexes and live retrieval. Coverage in credible publications can create exactly the kind of third-party corroboration these systems frequently draw on. Expect PR to matter more, not less, as this matures. I’m a huge advocate for PR as part of the marketing mix. I’ve seen firsthand the lift it can create. 

4. Get the technical basics right.

Clean HTML, schema markup where relevant, logical site structure, descriptive URLs. None of this is new. You just can’t skip it.

5. Publish original material.

If you give AI something unique — real data, original analysis, a case study with specific numbers — you give it a reason to cite you instead of a dozen interchangeable alternatives.

What to Skip

Skip the llms.txt hype (for now). We are all being told we need an llms.txt file, which is a proposed standard for telling AI crawlers what’s on your site. Ahrefs found almost no measurable evidence that llms.txt affected AI referral behaviour: 97% of llms.txt files received zero traffic in May 2026, and no major AI platform has committed to honouring the standard. It costs little to add and might matter later, but it’s a fourth-decimal-place optimization. Spend the hour rewriting your FAQ page in plain English instead. One cheap win that is measurable: freshness. Otterly reported a 61% increase in AI citations from adding the current year to page titles. Date your content, and update it when the data moves. (Case in point: this post.) 

Tools

This space is moving fast — faster than any category I’ve watched in twenty-five years of doing this. Since spring, the tooling has sorted into three tiers:

•    Check your existing stack first — if you already pay for Semrush or Ahrefs, you now have AI visibility tracking and don’t need a new subscription. Semrush’s AI Visibility Toolkit runs on an index of over 100 million prompts and includes a “prompt volume” metric — essentially search volume for AI questions. Ahrefs shipped Brand Radar for tracking AI mentions and citations. Neither is as deep as the specialists, but they’re the right starting point for most teams.

•    Dedicated monitoring — Otterly AI remains the best affordable entry point and the most widely adopted pure-play. AthenaHQ is strong on citation probability and still offers a free audit. Profound is the enterprise option, and its new FactCheck feature—monitoring when AI systems make inaccurate claims about your brand—points to where this category is going: from watching dashboards to correcting the record. And, obviously, you can hire a contractor like me to do it all for you!

•    Research — SparkToro for audience intelligence, including which AI platforms your audience uses and what they ask. Kevin Indig’s Growth Memo remains the best primary research in the field. And manual querying — prompting the AI systems the way your customers would — is still the single most instructive thing you can do. I do it weekly, and I keep finding things I didn’t expect.

No one has this fully figured out.

This “discipline” barely existed a year ago. The tooling is early. Best practices are still forming. I’m not entirely sure how to price GEO work in an audit yet, and the few times I’ve tried to forecast how long it takes to move the needle, I’ve been wrong in both directions.

What is clear is the behavioural shift. More people are using AI to shape decisions before they ever visit a website. That’s not theoretical. It’s happening in B2B research, in considered purchases, in category discovery. It happened to me when buying a sauna.

What This Means for Your Strategy

The sauna company that got my money didn’t “do GEO.” They answered questions clearly. They wrote like humans. They described their product honestly. They showed up consistently across the web.

In a traditional SEO model, they’d struggle. In an AI-mediated environment, they were the obvious answer.

You don't need a separate "GEO strategy." You need to evolve how you execute your existing one...

Start here: ask how your brand shows up in AI responses. Do it yourself, manually, today. Identify gaps in clarity, structure, and coverage. Fix what's already on your site before creating net new content. Most companies haven't done even that.

Or let me do it for you. I run a free AI visibility snapshot: your brand and a key competitor, tested across the questions your buyers actually ask, on the engines they actually use. You'll see exactly where you stand. Fair warning: the results are usually not what people expect.

→ Get your free AI visibility snapshot

Resources Worth Your Time

•       Rand Fishkin / SparkToro — sparktoro.com/blog

•       Kevin Indig / Growth Memo — growth-memo.com (start with ‘State of AI Search Optimization 2026’)

•       EMARKETER FAQ on GEO/AEO — emarketer.com (April 2026, balanced overview)

•       Conductor 2026 AEO/GEO Benchmarks — conductor.com/academy/aeo-geo-benchmarks-report

•    SparkToro — “In 2026, Less than One Third of Google Searches Still Send a Click” (June 2026) — sparktoro.com/blog.

Related reading from sequenceDM:

•       AI Marketing Integration: What Separates Results from Hype

•       AI Agents in Marketing: What They Are, What They Can Do, and Why Your B2B Team Needs a Plan

Gregory De Rocher is the founder of sequenceDM, a Vancouver-based B2B marketing consultancy. He founded sequenceDM a decade ago after 17 years in director-level marketing roles at Fidelity Investments Canada.