If a vision agent costs 45x more to serve than an API agent for the same task, what fraction of next year's marketing budget should you redirect from SEO to AEO?
That's a real question now. The cost asymmetry is structural. The traffic mix is shifting. And no one inside the marketing team has a defensible number to bring to the CFO.
This post tries to give one. Not a spreadsheet. A framework, with a starting band that you can argue with.
The cost data, briefly
The previous post in this sequence (The 45x Argument) walked through the Reflex benchmark in detail. Two agents, one task. Vision agent: 550,976 tokens, 53 steps, ~1,000 seconds. API agent: 12,151 tokens, 8 calls, 19.7 seconds. Cost ratio: 45x. (Reflex, April 2026)
That number isn't going down. Better models reduce error rates but they don't reduce step counts. A vision agent must render every intermediate screen to interpret the next move. The cost is paid by whoever runs the agent, which means: the buyer.
Now the budget question.
The buyer is paying to read your site
Every SEO budget assumes a free reader. A human visits your page, costs you some pixels and a database query, and that's it. The cost of reading lives on the visitor's side and never enters your spreadsheet.
Agent traffic doesn't work that way. The cost of reading is now an explicit line item for the buyer, and the buyer (or the platform routing the buyer) is going to make decisions accordingly. Sites that cost 45x to read get visited 45x less, or skipped, or ranked lower in whatever soft preference the agent applies.
Reframe AEO this way: it's not a marketing tactic. It's cost reduction for your buyer, charged back to you in lost rankings if you don't do it.
That makes the budget question feel different. You're not "investing in AI traffic." You're paying to keep your read-cost competitive. Different mental model, different number.
Where the average B2B SaaS team is today
Most B2B SaaS marketing budgets I see split roughly like this: 40-60% paid (SEM, LinkedIn, paid social), 20-40% organic SEO, 5-15% content production, 0-5% AEO if it shows up at all.
The AEO line is rounding error. That ratio made sense in 2024 because agent traffic was rounding error too. It doesn't make sense now, and it'll make less sense every quarter.
I'm not telling you to flip the ratio. I'm telling you the budget conversation hasn't happened yet at most companies, and that's the actual problem.
Three questions before you allocate anything
Before picking a percentage, get answers to these. They're the ones I ask Synlig clients first, because the answers vary 10x between companies in the same vertical.
1. What fraction of your inbound queries already involve an AI assistant?
Most teams have not measured this. The honest answer for a typical B2B SaaS in 2026 is "we don't know, but probably more than we think." Check your referrer data for anthropic.com, openai.com, perplexity.ai, gemini.google.com. Check whether your support inbox cites ChatGPT in user questions. The first time you measure, you'll be surprised which direction.
2. Of those AI-mediated queries, what fraction surface your brand vs. competitors?
Pick five high-intent buying queries in your category. Run each through ChatGPT, Claude, Perplexity, and Google AI Overview. Count: does your brand appear? In what position? In Synlig audits the typical pattern is asymmetric. One Norwegian client (Nordic Lithium) showed up zero times across eight relevant searches before we did the work. Their direct competitor showed up in all eight. That gap is rougher than the AEO-score gap suggests, because citation is winner-take-most.
3. What does the 12-month trajectory look like?
The relevant question isn't "how much agent traffic now." It's "how much by the time today's budget decision actually moves rankings." Citation graphs in AI assistants build slowly. Schema changes, fresh content, aggregator listings, and review presence don't show up in answers the same week you ship them. The lag is real and it's not yet well measured. Treat it like SEO lag in 2010: longer than your CFO will be comfortable with.
If those three answers come back as "small share, low presence, growing fast," you have a budget problem that won't fix itself.
The measurability trap
Here's the part nobody on the marketing team wants to hear. SEO is measurable on a 30-day cycle. Agent visibility is not, yet. So budget gravity flows toward the measurable, every time.
But measurability lag is not the same as outcome lag. By the time AEO has clean attribution dashboards, the citation graph is already locked in. The companies that shaped which sources LLMs cite during the 2026-2027 ramp will keep being cited because that's how reinforcement learning works on training data and retrieval indexes.
Waiting for measurability is what every loser in every platform shift does. It feels prudent. It loses.
A reasonable starting band
Reasonable, not validated. I don't have proprietary survey data on what high-performing B2B SaaS teams spend on AEO, because nobody has that data yet (the category is too young). What I have is the cost asymmetry, the trajectory, and the citation-graph mechanics.
Given those, the band I'd argue for B2B SaaS in 2026 is 10-25% of the digital marketing budget redirected toward AEO over the next 12 months, scaled by:
- How much of the buying journey already touches AI agents (varies wildly by vertical: dev tools high, niche compliance lower)
- How well your competitors are doing in AI assistant citations today (a competitor with 80% citation share is much harder to dislodge than one at 20%)
- How long the sales cycle is (long cycles = earlier investment pays back; short cycles = later is fine)
Treat 10% as the floor for any team where buyers use AI assistants weekly. Treat 25% as the ceiling unless you have specific evidence your category is ahead of the curve. If you spend 0% on AEO right now, going from 0 to 10% is the conversation. Going from 10 to 20 is next year's.
What we do at Synlig
Synlig Digital audits where your brand stands in AI assistant answers across 200+ buying queries, in 20+ categories, across 5 Norwegian cities. We give you a score (A-F), the gaps that explain the score, and a redirect plan.
Free check: sjekk.synligdigital.no
Paid analysis (4,900 NOK), full implementation (14,900 NOK), monthly monitoring: synligdigital.no
The budget redirect conversation is the actual blocker for most teams. The audit gives you the number to bring to it.
Previous post: The 45x Argument: Why Agent Economics Make AEO Non-Optional
Top comments (0)