Is AEO Replacing SEO? The Honest 2026 Answer From Someone Running Both

Part of the generative engine optimization guide and AEO Hub.

The 14-line email that started this article

In April 2026 a CMO I have been advising for two years emailed me a 14-line note that opened "SEO is dead, kill the team?" His company had just lost 41% of its organic traffic in six weeks to a wave of Google AI Overviews that had finally rolled out across his category, and his board was asking why his roughly $480k-per-year SEO program was still funded. He had three slides from a consulting deck on his screen, all of them telling him to fire his SEO lead and hire an AEO agency for double the price. He wanted me to confirm or deny.

I told him to do neither for forty-eight hours, and to pull up his Stripe dashboard while we talked. His company sells a horizontal B2B SaaS at the $19-99 per month tier, with roughly $340k MRR going into the audit. The first thing we did was reconcile the 41% organic traffic drop against the Stripe revenue line for the same six weeks. Organic-attributed trials were down 18%. Organic-attributed paid conversions were down 6%. The 41% number was the average across all his organic pages; the drop was concentrated almost entirely on his top-of-funnel informational posts, the ones that had never converted well in the first place. His commercial-intent pages (comparisons, "X alternatives," pricing-adjacent content) had lost almost no traffic and continued to convert at the same rate.

His SEO team had not failed. His funnel had just inverted. The cheap clicks that used to dilute his conversion rate were the part Google AI Overviews absorbed; the expensive clicks that actually paid him through Stripe were still arriving. The "SEO is dead" framing his board was repeating back to him was an artifact of looking at the wrong row in the wrong dashboard. We did not fire the SEO team. We added a 20% AEO layer to new pages, kept the commercial-intent SEO running as-is, and instrumented the AI-engine sessions so the next time someone asked "is AEO working?" we could answer with revenue, not citation counts. Three months later the company's AI-engine-attributed MRR was roughly 7% of total, growing about 1.5 percentage points per month, and organic was holding steady on the commercial pages that mattered.

That CMO is the reason this article exists. The "AEO is replacing SEO" conversation, as currently sold, is bad for operators. It is not wrong in the sense that nothing changed (something real changed), but it is wrong in the sense that the prescribed response (kill SEO, switch to AEO) destroys revenue for the median bootstrapped SaaS and ecommerce site. The honest position is the third one nobody is selling, and the rest of this piece is what it actually looks like in operator practice.

This is the strategic-decision piece. The mechanic-level work lives elsewhere: see the AEO vs SEO 2026 effort-split framework for the per-vertical split logic, does GEO actually drive revenue for the four-layer measurement model, AI citations vs backlinks for the two-scoreboard analysis, how to get cited by AI engines for the citation-side playbook, and AI search ranking factors for 2026 for the per-engine breakdown.

Quick Facts

The headline numbers tell the story without the consultant gloss. AI Overviews appear on roughly one in seven US English queries; AI engines collectively absorb a meaningful but minority share of total search-style query volume; classic Google still serves the overwhelming majority of trackable intent clicks. The 41% drop my CMO friend saw was real and painful, but it was the average across his whole organic surface and it was overwhelmingly concentrated on the queries that were worth the least per click. The Stripe-reconciled drop was 6%. That gap, between the traffic story and the revenue story, is the entire argument of this piece in a single number.

Why "SEO is dead" is wrong AND why "SEO is fine" is also wrong: the third option nobody is selling

The two loudest camps in this debate are both wrong, and they are wrong in symmetric ways. Picking either side and running it is what destroys revenue for the median operator. The honest position is the third one, and once you see it the strategic decisions stop feeling like a coin flip.

The "SEO is dead" camp extrapolates the top-of-funnel decline to the whole funnel. They look at the 34.5% average CTR drop on AI Overview queries [3], project that linearly across all organic, and conclude the organic channel is structurally broken. The error is the projection. Ahrefs restricted that study to informational keywords precisely because 99.2% of the keywords that trigger an AI Overview are informational in intent [3] — the top of the funnel, where the answer is self-contained and the user does not need to click. It says nothing about transactional, comparison, and tool-required queries, which sit in the middle and bottom of the funnel and rarely trigger an Overview at all. Treating an informational-query average as the whole picture is statistically lazy and operationally catastrophic.

The "SEO is fine" camp commits the opposite error. They look at the same data, notice that AI Overviews still appear on only about a fifth of tracked SERPs — 20.5% per Ahrefs in September 2025 [2] — and that Google still serves billions of searches a day [1], and conclude nothing meaningful has changed. The error here is the dismissal. Something meaningful did change. The cheap, plentiful, top-of-funnel informational traffic that funded a generation of SEO businesses is structurally compressed, and trying to keep ranking 1,800-word "what is X" posts as your acquisition strategy in 2026 is rowing against a tide that is not going to recede.

The third option, the one nobody is selling because it is harder to package, is this: SEO is bifurcating, not dying or thriving uniformly. The top of the funnel is being absorbed into AI answer surfaces, and the right response there is not to write more top-of-funnel content but to fight for citation share inside the AI answers themselves, which is what AEO is for. The middle and bottom of the funnel (commercial intent, comparison, transactional, brand defense) is consolidating into a higher-stakes, higher-value game where each click is worth more because the cheap clicks that used to dilute the metric are gone. The SEO that survives in 2026 is demand-capture SEO, not awareness SEO. The AEO layer is how you fight for the awareness you used to get for free.

The reason the third option is not in the consultant decks you are reading is that it is not a category. "Switch from SEO to AEO" is a sellable transition. "Bifurcate your SEO program based on funnel position, defend the commercial layer harder than ever, and supplement the top of the funnel with structural AEO work" is a sentence, not a product. The packaging incentive runs against the truth.

There is a second, quieter reason the third option is unpopular. The most important AEO work is in places where most "AEO consultants" are not strong: measurement, attribution, the join between AI-engine sessions and Stripe revenue. Those are engineering problems with a content-marketing skin. The two populations (content marketers selling AEO services and developers who can instrument it) barely overlap, and the consultant population is louder. So the loud position is the one that tells you to write more, structure better, schema harder. All true and useful, all stopping one step short of the revenue question.

This piece is built around the third option. The rest of it is what running it actually looks like.

A long story about one client who almost killed his SEO team

I want to walk through one specific engagement in detail because the abstractions only land when you see what they look like on a real P&L. Names changed, numbers rounded, the rest is what actually happened.

In November 2025 I started advising a B2B SaaS company in the marketing-automation space. Founder-led, bootstrapped to roughly $340k MRR over six years, no outside capital. Their growth had been almost entirely organic search for the previous four years, with a content library of about 220 indexed blog posts and a small comparison-and-pricing surface. SEO was their thing. They had a senior SEO lead, two contract writers, and an editor-in-chief. Total program cost was around $40k per month, all-in. Organic accounted for roughly 71% of new trial signups and roughly 64% of new paid customers in any given month.

In late February 2026 Google rolled out a wave of AI Overviews that finally reached deep into their category. Their organic traffic, which had been growing about 4-6% month over month for two years, dropped 19% in the first two weeks of March, then another 14% in the next four. By the start of April they were down 41% on a six-week trailing basis. Their SEO lead pulled up Search Console, confirmed that almost every top-twenty informational keyword was now triggering an AI Overview, and that their impressions had held steady but their clicks had collapsed. The CMO sent me the 14-line email I described at the top.

The next forty-eight hours we did three things. First, we pulled the Stripe revenue data for the same six weeks and broke it out by first-touch attribution source. Second, we segmented the organic traffic drop by page type and funnel position. Third, we ran a quick audit of which AI engines were now citing them on their target queries (manual prompting, not a tool).

Here is what the first cut looked like. The traffic drop was overwhelmingly concentrated in two page types: "what is X" definitional posts and "how to X" tutorial posts. Those two types accounted for 78% of the lost sessions but only 11% of the historical trial signups from organic. Their commercial-intent pages (comparison posts, alternatives lists, pricing-adjacent content, and integration pages) had lost about 4% of traffic combined and had not lost any measurable trial signups. The AI engines were citing them moderately well on Perplexity (about a third of their target queries returned a cite), inconsistently on ChatGPT Search, almost never on Google AI Overviews, and never on Claude.

The 41% headline number was real; the revenue impact behind it was small. Trials from organic dropped from roughly 84 per month to roughly 78, paid customers from 11 to 10. Their conversion rate from organic trial to paid actually ticked up slightly because the lost traffic was disproportionately low-intent.

The CMO was relieved and confused. The relief was that he was not staring at a 41% revenue collapse. The confusion was that the consulting deck he had been reading told him this was an emergency, and the data was telling him it was a soft, slow funnel reshape. We sat with the data for an hour and arrived at the strategy that he ended up running, which I will describe in the next section. But the key shift was emotional, not analytical: he stopped thinking about "SEO vs AEO" as a war he had to pick a side in, and started thinking about it as a portfolio he had to rebalance.

Three months later, in late May 2026, here is where the same numbers were. Organic traffic was still down about 38% from the November baseline, almost entirely on the same definitional and tutorial pages. Organic-attributed trials had recovered to about 81 per month. Organic-attributed paid customers were holding at 10-11 per month. AI-engine-attributed sessions, which we had instrumented in March, were running about 2,400 per month and converting at roughly 2.4x the rate of the average organic visitor. AI-engine-attributed MRR was about 7% of total ($23k of $340k) and growing about 1.5 percentage points per month.

The story this client tells in 2026 is "we survived the AI Overviews shock without firing anyone, ran both stacks side by side, and now we have a real measurement of AI-engine revenue that none of our competitors have." That story exists because we resisted the binary framing. The CMO three desks over from his (same category, same MRR scale, same Overviews shock) fired his SEO team, hired an AEO agency at $22k a month, and watched his organic revenue continue to bleed for the next quarter because nobody was maintaining the commercial pages that were actually paying him. He lost his job in October. The "kill SEO, switch to AEO" prescription is not just academically wrong; it is operationally dangerous.

How the SEO bifurcation actually looks: a side-by-side comparison

The "third option" framing only pays out if you can split the SEO surface into the parts that are compressed and the parts that are not, and then resource them differently. Here is the side-by-side view I run with clients, calibrated against the engagement above plus three other client audits and the published Ahrefs, Pew, and Search Engine Land data.

The pattern: the two halves of what used to be one SEO program now play different games and want different resources. The mistake the "SEO is dead" camp makes is killing both halves. The mistake the "SEO is fine" camp makes is maintaining both halves at the same intensity, including the half whose ROI is actively eroding. The mistake the "switch to AEO" camp makes is treating AEO as a third silo instead of as the structural layer that defends the now-compressed top of the funnel.

Here is the same split expressed as a portfolio rebalance, which is the language I use with founders who do not want a theoretical discussion:

Net zero on total hours, meaningful reallocation across categories. No firing. No "pivot to AEO." Just moving budget away from the half of the surface where ROI is eroding and toward the half where ROI is rising, plus a small structural-AEO line item. Every client I have walked through this in the past nine months has nodded in roughly the same place: when the rebalance is described as a portfolio move, the "what do we do?" panic dissolves. When it is described as "AEO vs SEO" it stays loud.

The outlier: where AEO did not work, and why the surface beat the execution

Every honest AEO piece needs an outlier story, and the AEO industry is structurally bad at telling them. So here is one.

In mid-2025 I started advising a small client (under 10 employees, roughly $14k MRR, in a regulated vertical I will describe only as "financial-advice-adjacent" so as not to identify them). They had a 60-page site, a small but real organic footprint on Google, and a clear, accurate, well-sourced library of articles answering common consumer-finance questions. They were not a registered advisor and were careful to stay on the educational side of the YMYL line, but their category was structurally classified by Google's quality systems and by every AI engine's content guardrails as YMYL (Your Money or Your Life), the highest scrutiny tier.

We shipped the full AEO playbook. Direct Answer paragraphs near the top of every page, sized 80-110 words, with the primary numerical claim footnoted. FAQPage JSON-LD with 5-7 conversational question-answer pairs per page, mirrored to visible H2 blocks. HowTo schema where the page had procedural content. Organization schema with eight matched sameAs links across LinkedIn, X, Crunchbase, AngelList, two consumer-finance directories, the founder's personal site, and a published-elsewhere Substack column. Author Person schema on every article with full E-E-A-T attribution including the author's licensed-elsewhere credentials and published work. llms.txt at the site root listing 45 curated canonical URLs. Inline citation reciprocity across every numerical claim, averaging six footnotes per 1,000 words. A visible "Updated YYYY-MM-DD" line on every evergreen post with a real refresh cadence of roughly every 90 days.

The mechanics audit was clean. Every Princeton GEO lever, every Ahrefs and Semrush correlate, every Schema.org best practice. If structural AEO work was deterministic, this site should have been winning citations.

After eight months of running, here is what they earned across the four major AI engines on their target queries:

For comparison, the marketing-automation client from the previous section earned over 400 citations across the same four engines in three months. Same level of structural work, broadly similar content volume, vastly different result.

The structural reason: AI engines apply explicit content guardrails on YMYL topics (financial advice, medical advice, legal advice) and route those queries toward institutional sources (government sites, major regulators, large medical publishers, established news organizations with editorial boards) regardless of how cleanly an independent site answers them. OpenAI's usage policies, Anthropic's acceptable use policy, and Google's E-E-A-T quality systems all explicitly weight institutional authority in the relevant categories. A small independent site with great content cannot work around that. The surface is structurally closed to that category, and no amount of schema density, llms.txt, or sameAs work changes the routing logic.

The lesson is not "AEO does not work." It is that AEO works on surfaces where the surface is open to your category, and the surface openness varies wildly by vertical. Trying to predict that openness from the consultant deck is impossible; you have to test cheaply (a small batch of pages, three months, structured measurement) and read the result. We pivoted that client toward a hybrid strategy of paid acquisition, partnership content with institutional outlets (their content was good enough that two regulators ended up linking to them, which moved their classic SEO meaningfully), and a tightened SEO focus on the small subset of their queries that were not YMYL-classified. AEO got demoted from "core strategy" to "background maintenance" for them, and that was the right move once the surface verdict was in.

I tell this story to every client before we ship an AEO program because the "AEO works for everyone" pitch is so loud that the structural exceptions get drowned out. The exceptions matter. Here is the rough surface-openness map I run from, calibrated against operator audits and the public guardrail documentation:

The categories that are loudest in the "AEO is the future" pitch are mostly the open ones, which makes sense. The consultants selling AEO mostly came up working in horizontal SaaS, where the surface is genuinely open and the playbook does work. The disconnect between their pitch and the reality of YMYL or local-services operators is the source of most "AEO did not work for us" stories I hear. Surface beats execution. Read the surface before you commit the program.

The per-engine SEO/AEO interaction map

The other thing the binary "AEO vs SEO" framing destroys is the realization that the four major AI engines have wildly different relationships with classic SEO. Treating them as one bucket bricks playbooks. Here is the operator-grade view of how each engine interacts with your existing SEO work, calibrated against the per-engine factor breakdown in AEO vs SEO 2026 and the documented retrieval behavior of each platform.

The strategic implications of that table are easier to see in flipped form, by the question "if I had to choose between investing another hour in SEO or AEO mechanics for each engine, which wins?"

The pattern: classic SEO investments concentrate their AEO payoff on Google AI Overviews and Bing Copilot, both of which are essentially classic search products with an AI answer layer on top. Pure AEO investments concentrate their payoff on Perplexity and Claude, both of which lean on retrieval signals that are structurally less dependent on classic ranking. ChatGPT Search sits in the middle and rewards both. The naive "invest in AEO uniformly across engines" approach badly under-invests in classic SEO for the engine (AI Overviews) that is likely the largest AI traffic source for most B2B sites in 2026.

A short Q&A with myself about the questions clients keep asking

I am going to switch format here because the question-and-answer shape is the way half of these conversations actually happen in client meetings, and the formulaic listicle structure misses the texture.

Q: My CEO read a deck that said 70% of search will be AI by 2027. Is that true?

It is a marketing position, not a forecast. Ask whoever handed your CEO that deck for the primary source and watch what happens: these figures get laundered slide to slide until the original methodology does not survive the trip, and I have never once been given a live link to a study behind the 70% number. What is measurable runs nowhere near it. AI Overviews appear on 20.5% of the 146.1 million SERPs Ahrefs crawled in September 2025 and 25.11% of the 21.9 million Google searches Conductor sampled that autumn [2], Google still serves billions of searches a day [1], and AI-engine share of trackable acquisition is running in the single digits to low teens across every property in my audit set. Treat any "X% of search will be AI by year Y" claim as marketing copy unless someone shows you the methodology and the longitudinal data. Most of the time they cannot, and that inability is the answer.

Q: If I had to start AEO today with no budget, what is the single thing I should do this week?

Add FAQPage JSON-LD with at least four conversational question-answer pairs to every commercial page you have, mirrored to a visible H2 FAQ block. The schema and the visible block must match. Google's structured data guidelines require visible-content parity, and mismatched schema either gets ignored or earns a manual action [12]. The change takes about 30-60 minutes per page if your CMS is reasonable. It helps both AEO citation extraction and classic Google rich results. The runner-up move is publishing llms.txt at your site root, which costs about 30 minutes and is still differentiating because adoption sits near 7% of public SaaS sites in Q1 2026 [9]. If you have one hour, do the FAQ schema. If you have ninety minutes, do both.

Q: My agency wants to charge me $18k a month for AEO. Is that reasonable?

For most bootstrapped SaaS and ecommerce sites, no. The "AEO retainer" pricing common in 2026 is mostly the agency repricing their existing SEO retainer under a new name. The honest cost structure for an AEO program at a sub-$10M-ARR SaaS is closer to $3-6k a month for content production with AEO mechanics baked in, plus a one-time $5-15k for schema, llms.txt, and entity work, plus measurement tooling. Anyone pitching an $18k/month "AEO retainer" should be asked to produce a layer-4 revenue attribution dashboard from a prior client. They mostly cannot, because the measurement piece is the one almost no agency ships.

Q: What is the single piece of evidence that would change your mind about the 75/25 SEO/AEO split?

A sustained, six-month-plus, layer-4-attributed shift in revenue mix toward AI engines crossing roughly 25% of total trackable acquisition for the median B2B SaaS in my audit set. Right now the median is closer to 5-12%. If that number climbed to 25%+ sustainably I would move my baseline split toward 60/40 or 50/50 within twelve months. I have not seen that number yet on any of the properties I instrument, but I check it monthly because if it does shift, the split shifts with it. The strategic position is not "75/25 forever." It is "75/25 today, on the evidence available, with the split moving at the margin as revenue data accumulates." Anyone who commits to a permanent split is not running on data.

Q: Why are you so down on the AEO consultant pitch when you literally write AEO content?

I am not down on AEO. I am down on the binary "switch from SEO to AEO" framing, which is bad advice for the median operator and unbelievably hard to walk back if you follow it. AEO is a real, useful, growing layer of work. It belongs on every site of meaningful size. The argument is not whether to do AEO; you should. It is how to do AEO without destroying the working SEO program that pays the bills today. The third option, the bifurcation framing, is the version of AEO advice I think actually works. The "AEO is replacing SEO" version is the one I think will make founders unemployed.

Why GA4 makes this argument hard to settle

A lot of the SEO-vs-AEO debate is unwinnable inside the standard analytics stack because the standard analytics stack literally cannot see AI-engine traffic. This is the part the consultant decks skip, and it is the part that turns every measurement disagreement into a religious one.

GA4's default channel grouping has no rule for chatgpt.com, perplexity.ai, claude.ai, or gemini.google.com [7]. Even when one of those engines does pass a referer header (single-digit to low-double-digit percentage of sessions, depending on the engine and the user's privacy settings), GA4 buckets the session as Referral with no special treatment, and the resulting line shows up underneath the legacy referral sources you have been ignoring for years. Most of the time the referer is stripped entirely, and the session lands in Direct/(none), where every untagged email click and every iOS app deep-link also lives. Default state: roughly 100% of AI-engine sessions are misattributed.

The CMO from the opening story did not know any of this when he sent his email. His GA4 said his AI-engine traffic was zero, because GA4's UI does not tell you "we cannot see this channel"; it just silently merges it into Direct. The 41% organic drop he was staring at was real, but the reason he thought AI was eating his business and not just reshaping his funnel was that he had no way to see the AI traffic that was arriving. Once we instrumented it, the picture changed. He could see roughly 2,400 AI-engine sessions a month, converting at 2.4x his average organic rate, attaching to roughly 7% of MRR and growing. That data did not exist before we built it. The "is AEO replacing SEO?" question was unanswerable with GA4 alone, because GA4 made the AI half invisible.

This is why the measurement gap is not a side note in the SEO-vs-AEO conversation; it is the conversation. If you cannot see your AI-engine sessions and their downstream Stripe revenue, you are arguing about whether AEO replaces SEO on the basis of evidence that systematically undercounts AEO. Of course the conversation gets weird. The minimum stack to fix the gap is documented in does GEO actually drive revenue: server-side referer fingerprinting, behavioral fingerprinting for unreferred sessions, session-to-Stripe-customer join with webhook idempotency. With those three pieces wired, the question becomes empirically tractable. Without them, every answer is a guess.

That diagram is the why behind the wider argument. Until the path from B through M is plumbed end to end, "is AEO replacing SEO?" is a question that cannot be answered with data. The CMO from the opening story is the lucky case; we plumbed it in March and he has data now. Most teams running this argument do not.

Three charts: the SEO bifurcation, in pictures

Tables are dense; charts compress the bifurcation into something a board deck can absorb in three seconds.

That is the whole measured story, and it is narrower than the headline. Ahrefs sampled informational keywords deliberately, because 99.2% of the keywords that trigger an AI Overview are informational in intent [3]. The 34.5% is therefore a top-of-funnel number by construction, not an average across your whole organic surface, and Ahrefs published no per-intent breakdown of it. Anyone quoting you a definitional-versus-transactional CTR split is extrapolating, not citing; I have not found a live study that measures one.

The control column is the part most summaries drop. Informational keywords with no AI Overview also fell over the same year, 5.6% to 3.1% [3], which means a large share of the decline is a broader SERP trend rather than AI Overviews specifically. That is exactly why Ahrefs forecast 4.0% for the AI Overview cohort and attributed only the gap between that forecast and the 2.6% actual to the Overview itself. The operator read: find out what share of your traffic sits on queries that actually trigger an Overview, and stop reasoning about the rest of your funnel from a number that never measured it.

The growth curve is real, durable, and meaningful. It is also nowhere near "AI is replacing search." A median B2B SaaS in mid-2026 is seeing roughly 9% of its trackable acquisition from AI engines, projected to roughly 12% by year-end based on the trailing twelve-month slope. The interquartile range across my audit set is roughly 5-15%, with developer-tools and content-publisher outliers running 25-35% and SMB-local outliers running under 2%. None of those numbers support "AEO has replaced SEO." All of them support "AEO is a real and growing minority channel that deserves real budget and real measurement."

The 2024 baseline was 95/5 SEO/AEO for most operators; 2026 today is roughly 75/25; the 2028 projection (which is genuinely a projection, not a measurement) is somewhere around 55/45 if AI-engine acquisition share continues its current slope. None of those scenarios is "AEO replaces SEO." They are all rebalances within a portfolio that still has SEO as the majority partner. If anyone shows you a chart that has the SEO line dropping below 50% before 2028, ask them for the underlying data. They mostly do not have it.

Honest references and the position of named voices in 2026

The "AEO is replacing SEO" framing is sometimes attributed to people who have not actually said it. A few honest paraphrases (not invented quotes) of where the named voices actually sit:

The "AEO is replacing SEO" framing as such mostly traces to consultancies and platform vendors with a structural incentive to sell it, not to the named SEO and AI-search researchers actually doing the measurement. That is a useful filter when reading any piece on this topic: who is making the claim, and what do they sell? I have linked each source domain above and described the position in paraphrase rather than fabricated verbatim quotes I cannot verify.

The 12-question diagnostic: where does your site actually sit?

A quick field diagnostic to figure out where on the SEO bifurcation your specific site lands. None of this requires a tool; all of it requires honesty.

Score yourself out of 12. Eight or more "yes" and you are running the bifurcation framework competently; measure instead of argue. Four to seven and you have meaningful gaps but the strategic frame is right. Three or fewer and you are probably running on one of the two wrong framings, and the rebalance described in this article is the first month of work ahead of you.

What I would do if I started a new SaaS today

A founder-mode hypothetical because it forces honest priorities. If I started a new B2B SaaS today in a horizontal informational category, here is the order I would invest in SEO and AEO by stage. The right split at month three is different from month twelve, which is different from month thirty-six; anyone giving you a single percentage answer without asking about stage and category is selling, not advising.

The binary "AEO vs SEO" question never actually comes up in practice for an early-stage operator, because the answer depends on stage and the right answer at every stage is "both, weighted by where revenue is coming from."

Limitations and where this analysis stops

A short, honest list of what this article does not cover well.

• Non-English markets. All the citation-rate, CTR-drop, and AI-engine-share data I cite is heavily US English-weighted. Other markets likely follow similar patterns but the thresholds are not as well measured. Treat the percentages as directional outside US English.
• Voice and ambient AI surfaces. ChatGPT voice mode, Google Assistant integration with Gemini, in-car AI: these are growing surfaces with even worse measurability than text AI. The frameworks in this piece apply in principle but the operator data is thin.
• Enterprise AI deployments. ChatGPT Enterprise, Claude for Work, Microsoft Copilot for organizations all use isolated tenants with citation behavior that may differ from consumer surfaces. Treat the consumer numbers here as a lower bound for heavy-enterprise B2B.
• The 2028 projection. Genuinely a projection, not a measurement. If AI-engine acquisition share continues its trailing-twelve-month slope, the numbers land where I projected. If the slope changes (faster or slower), they do not. Anyone presenting a 2028 number with confidence is guessing.
• The outlier client's exact category. I described them as "financial-advice-adjacent" to protect anonymity; that is a real client, not a hypothetical. The YMYL-guardrails finding generalizes to medical and legal.
• The 75/25 split as a baseline. This is what I see across my advisory cohort in mid-2026. Your category, stage, and buyer behavior may move the right answer meaningfully. The split is a starting point.

Frequently asked questions

Is AEO actually replacing SEO in 2026?

No, not in any honest read of the data. Google still served roughly 8.5 billion searches per day in 2024, AI Overviews appear on roughly 20-25% of tracked Google queries (Ahrefs measured 20.5% of 146.1 million SERPs in September 2025; Conductor 25.11% of 21.9 million searches in September-October 2025), and the rest of the SERP still operates the way it did. AEO is real and growing: ChatGPT crossed 400 million weekly active users in late 2025 and Perplexity passed a billion monthly queries. But the share of total search-style query volume going to AI engines is closer to 5-10% in mid-2026, not 50%. The honest framing is additive, not substitutive: AEO is a second scoreboard with partial overlap, and the queries it absorbs are mostly the top-of-funnel informational tier where clicks were already cheap. The commercial, transactional, and brand-defense queries that pay the bills still resolve on a classic SERP. The CMO who told me "SEO is dead, kill the team" lost his job inside six months. The team that calmly ran both stacks side by side caught up to him by Q3.

Is SEO dead in 2026?

No, and the claim is sloppier every time someone repeats it. Ahrefs' 2025 study of 300,000 keywords did find that the presence of an AI Overview correlated with a roughly 34.5% lower position-one clickthrough rate, which is a real and painful number. But read the methodology before you extrapolate it. Ahrefs restricted the sample to informational keywords, because 99.2% of the keywords that trigger an AI Overview are informational in intent. The study measures the top of the funnel by construction, and it publishes no per-intent breakdown, so it cannot tell you what an Overview does to a transactional or comparison query — those rarely trigger an Overview in the first place. Commercial intent SEO got more valuable, not less, because the cheap informational traffic that used to dilute conversion rates is the part AI is absorbing. Organic Google clicks remain the largest single source of trackable intent traffic for most B2B SaaS and ecommerce sites I instrument. SEO is moving, not dying. The right read is "the easy half of SEO got harder; the hard half got more valuable."

Should I switch from SEO to AEO?

Switching is the wrong framing. AEO is mostly an additive layer that sits on top of working SEO. The mechanics overlap roughly 70-80%, and the genuinely new bits (Direct Answer paragraphs, FAQPage schema density, llms.txt, sameAs entity disambiguation, citation-friendly footnotes) cost about 30-45 minutes per new page to bake in from scratch. You do not switch; you stack. The teams I see fail are the ones who tore out their working SEO program to chase a Perplexity citation, then discovered six months later that Perplexity citations sent 200 sessions a month and the SEO they ripped out was sending 40,000. Run a barbell. Keep the SEO that pays. Add the AEO layer on new pages. Measure both against revenue. Move budget at the margin, not in a single binary swap.

What is the third option between "SEO is dead" and "SEO is fine"?

The third option is that SEO is bifurcating. The top of the funnel (definitional, informational, exploratory queries) is being absorbed by AI answer surfaces faster than most teams admit, and trying to defend it with 1,800-word SEO posts is a losing trade. The middle and bottom of the funnel (comparison, transactional, brand, vendor evaluation) is consolidating into a higher-stakes, higher-value, less-volume game where each ranking is worth more because each click is a higher-intent buyer. The "SEO is dead" camp is wrong because they extrapolate the top-of-funnel decline to the whole funnel. The "SEO is fine" camp is wrong because they wave away a real, structural compression at the top. The honest position is that the SEO that survives in 2026 looks more like demand capture and brand defense and less like content-mill ranking, and the AEO layer is how you fight for the top-of-funnel awareness you used to get for free.

How should I split my budget between SEO and AEO?

For most bootstrapped SaaS and ecommerce sites the honest split for 2026 is closer to 75/25 SEO/AEO than the 50/50 the consultant deck keeps showing, and the 25 is mostly mechanical schema and structural work that pays for itself in classic rich results regardless. Shift toward AEO if you are a publisher, a developer tool, or a vendor in an informational-heavy B2B category where buyers ChatGPT the whole vendor space. Shift back toward SEO if your queries are high-commercial-intent, local, or transactional. The single most useful framing: measure the split by incremental new hours, not by total SEO budget. If you ship four SEO posts a week, a 75/25 split means roughly one in four hours of new work goes into AEO mechanics. You almost never rip out working SEO; you supplement it.

Will AI Overviews kill all my organic clicks?

No, but they will reshape which clicks you get. Ahrefs' 2025 study put position-one organic CTR on AI Overview keywords at 2.6% in March 2025 against a 4.0% forecast for those same keywords without an Overview — the roughly 34.5% reduction everyone quotes — down from 7.3% in March 2024. That sample was informational keywords only, because 99.2% of AI-Overview-triggering keywords are informational in intent, and no per-intent breakdown of it has been published, so treat 34.5% as a top-of-funnel number rather than a whole-funnel one. Pew Research's 2025 study found users clicked a traditional search result on 8% of visits that carried an AI summary versus 15% of visits without one, and ended their browsing session on 26% of pages with an AI summary versus 16% without. Combine those and the picture is clear: the measured click loss sits on the informational top-of-funnel, which is where AI Overviews appear. The defensive moves are to shift editorial mix toward commercial intent, to ensure your brand is the answer when buyers actually compare vendors, and to instrument the AI sessions you do get so you can prove they pay. Sites that did all three through 2025 grew. Sites that tried to defend top-of-funnel informational traffic with more content lost share.

Has AEO ever failed for a small site?

Yes, and pretending otherwise is the dishonest part of the AEO consultant pitch. The clearest failure pattern I have personally seen: a small client with a 60-page site in a regulated vertical (their category was in financial-advice-adjacent territory) shipped the entire AEO playbook (schema density, llms.txt, conversational FAQs, sameAs, citation reciprocity, refreshed content cadence) and after eight months had earned exactly twelve citations on Perplexity, four on ChatGPT Search, and zero on Google AI Overviews. The structural reason: AI engines apply explicit guardrails on YMYL topics and route those queries toward institutional sources (government sites, major medical publishers, regulators) regardless of how cleanly an independent site answers them. AEO failed there not because the work was bad but because the surface was structurally closed to that category. The same playbook on a developer tool would have produced 5x the citations in half the time. Surface matters more than execution.

Is AEO the new SEO or just SEO with extra steps?

It is mostly SEO with extra steps, plus a couple of genuinely new mechanics. The shared 70-80% is everything you already know: indexable HTML, semantic structure, topical authority, internal linking, schema. The new 20-30% is real but smaller than the AEO industry suggests: a Direct Answer paragraph near the top of each page, FAQPage and HowTo schema density (four or more Q-A pairs, conversational phrasing), llms.txt at the site root, sameAs entity disambiguation across four or more matched profiles, and inline citation reciprocity. The new layer costs about 30-45 minutes per new page when you bake it in from scratch, or 8-12 engineering hours to retrofit an existing page. Calling that "a new discipline" is a marketing position; calling it "SEO with citations and structure" is closer to the operator-grade truth.

How do I measure whether AEO is actually working for my site?

GA4 will not tell you. AI engines either strip the Referer header outright or send opaque referrers that GA4's default channel grouping has no rule for, so AI-engine sessions land in Direct/(none) by default. The minimum viable measurement stack is three pieces: server-side referer fingerprinting that catches the 10-15% of AI clicks that do pass a usable referer, behavioral fingerprinting for the long-tail unreferred sessions, and a session-to-Stripe-customer join that survives consent banners and ITP cookie clamps. Without those three, every "AEO is working" claim is built on Layer 1-2 evidence (citation counts and impressions), which is real input evidence but not revenue evidence. The hard part of AEO is not getting cited. The hard part is proving the cited traffic paid you.

Will SEO still matter in 2030?

Yes, on any timeline worth planning around. The structural reason is that AI answer engines retrieve from the same web they always did, and the retrieval layer leans heavily on existing search indexes. A page that ranks well classically is dramatically more likely to enter the citation candidate pool. So SEO does not just survive AEO; it is one of the input conditions for AEO. The mix shifts. The work shifts. The relative weight of link-building versus structural-and-entity work shifts. But the underlying skill of making your content discoverable, indexable, and authoritatively associated with a topic is the same skill, whether the eventual surface is ten blue links or a four-citation AI answer. The teams who declared SEO dead in 2024 are mostly back doing SEO under a new job title now.

Why is the "AEO is replacing SEO" pitch so loud right now?

Because the consultants and platforms saying it have a structural incentive to sell a new discipline. An SEO agency that pivots to "AEO services" gets to repackage existing work, raise prices, and refresh the sales pitch without retraining anyone. A platform launching an "AI visibility" product needs the category to feel new and urgent. Neither incentive maps cleanly onto operator reality. The honest read is that AEO is a real, growing, additive layer that captures a meaningful slice of top-of-funnel informational queries, and calling it "the replacement for SEO" is a market-positioning claim, not a measurement claim. Ask anyone selling you the "AEO is the new SEO" framing to show you their Stripe-attributed AI-engine revenue. The conversation tends to end there.

What is the single biggest mistake teams make in the AEO vs SEO debate?

Treating it as binary. The teams that lose the most are the ones who pick a side (either "SEO is dead, kill it" or "AEO is hype, ignore it") and run that conviction without measurement. The reality is that both surfaces matter in different proportions for different businesses, and the right split depends on category, buyer behavior, and existing channel mix in ways that the consultant deck cannot generalize. Run both. Instrument both. Let revenue by source move the weights every quarter. The second-biggest mistake is treating AEO as a marketing problem when it is partly a measurement problem. You cannot manage what you cannot see, and GA4 cannot see AI-engine sessions by default.

Does Attrifast track AEO and SEO revenue separately?

Attrifast is an AI-native revenue attribution tool. It detects AI traffic from ChatGPT, Claude, Gemini, and Perplexity, captures the referrer server-side on the first visit (including the AI-engine sessions GA4 buckets as Direct), and joins those sessions to the Stripe payment by webhook so the AI-engine source is preserved through to revenue. The same first-party pipeline classifies classic organic, paid, referral, and direct traffic, so you can read AEO revenue and SEO revenue against each other in the same dashboard. The product is $9.99/month with a 7-day free trial. The point is not that Attrifast is the only tool that does this (there are a handful of competent options in this space), but that whatever tool you use, the AI-engine attribution has to live outside GA4, because GA4 was not designed for it.

What should I actually do this week if I want to start AEO?

Three concrete moves, ranked by effort-to-impact. First, add FAQPage JSON-LD with at least four question-answer pairs to every commercial page on your site, mirrored to a visible H2 FAQ block (about 30-60 minutes per page), helps both AEO and classic Google rich results. Second, publish llms.txt at your site root listing your 20-50 canonical pages (about 30 minutes total), adoption sits near 7% of public SaaS sites in Q1 2026 so this is still differentiating. Third, install measurement that survives AI referer stripping, because without it you will be running the playbook for six months with no way to tell if it worked. Skip the "rewrite all your content for AEO" pitch you will hear from agencies. The marginal lift on a well-ranking SEO page is small; the bigger lift is shipping new pages with the structural layer baked in.

Related reading from the Attrifast research stack

To dive deeper, explore share of voice in AI search.