Avinash Kaushik Says Cut Your SEO Retainer 25-75%. First, Own Your Data.
Avinash Kaushik, a 16-year Google veteran now chief strategy officer at Human Made Machine, told marketers this week to renegotiate agency contracts for 25% to 75% lower fees. His argument: AI and the ad platforms now handle the account structuring, bidding, and reporting that used to fill a retainer. The catch is that you need to own your own data before you have the standing to ask.
What Kaushik is actually repricing
The headline number reads like another "SEO is dead" story. It isn't. Kaushik published this in his Marketing Analytics Intersect newsletter under the title "Pay Less, Grow More," and the model underneath it is far more specific than the discount. He splits a typical retainer into work the platforms have quietly absorbed and work that still needs a human. The cut only applies to the first bucket. The second bucket, he argues, should actually cost you more.
That distinction matters, because most renegotiations I have watched go sideways start from the wrong premise. A client reads "cut fees 50%" and walks in asking for a flat 50% off the whole scope. The agency defends everything. Nobody moves, and six weeks later the contract renews unchanged. Kaushik's framing sidesteps that stalemate by naming which line items are supposed to shrink and which are supposed to grow.
The commoditized 80%, and why he is right about it
Kaushik estimates that account architecture, keyword and audience structuring, and campaign build-out (together roughly a fifth of a typical contract) can shrink by close to 80% now that platform algorithms handle segmentation and targeting better than a human team does. Manual bid and pacing adjustments land in the same pile. So does weekly reporting, which he figures can drop about 60% because the dashboards more or less assemble themselves now.
Here is the part worth sitting with. He isn't calling this work worthless. He is saying you are paying a senior rate for tasks a model now does in the background, and the platform handing you that model is the same one selling you the ad inventory. It is a bit like keeping a full-time switchboard operator on payroll after the phone system went digital. The task still technically happens. Nobody should be billing you senior strategist hours for it.
The platform selling you the automation is the platform selling you the ad inventory. Sit with that one.
The 15-25% that got more expensive, not cheaper
The mirror image of the cut is a raise. Kaushik puts 15% to 25% fee growth on genuinely new work: entity building, content architecture for AI Mode and AI Overviews, and GEO (generative engine optimization) strategy. This is the work that decides whether a model cites you at all, which is a different job than ranking a page. Kaushik has been making a version of this case for a while, most bluntly in his "Bye SEO, Hello AEO" essay.
Why it is suddenly worth more comes down to the traffic math shifting underneath everyone. The Pew Research Center found users click a traditional search result in 8% of searches that show an AI summary, versus 15% of searches without one. Clicks on the links inside the summary itself: about 1% of visits. Google disputes the methodology, and it is fair to flag that Pew's sample was 900 US adults, not the whole web. But the direction is not really in question. We covered a separate randomized test that measured a 39.8% click drop on affected queries. When the blue link delivers roughly half the clicks it used to, ranking first is worth less and getting named in the answer is worth more. That reprices the skill, not just the retainer.
The awkward part is that charging premium rates for GEO work runs straight into a measurement problem. Citation tracking across ChatGPT, Gemini, and AI Overviews is fragmented, and most tools only watch one surface at a time. We dug into how ChatGPT alone runs four separate search pipelines, which means a GEO report built on a single tracker is probably showing you a slice of where you actually get cited. So the work is genuinely harder and worth more, but the reporting that proves it isn't quite there yet. Price the work, discount the certainty.
Own your data first, or you have nothing to negotiate with
Before any of the fee talk, Kaushik puts one prerequisite, and honestly it is the actual homework here: own your data. GA4 admin access, Search Console, server log files, and whatever AI citation tracking you run all need to sit in your accounts, not your agency's. If the agency owns your GA4 property and your Search Console verification, you can't credibly threaten to leave, and you can't independently check whether a fee cut quietly hurt performance.
The 30-minute version: open GA4, go to Admin, Account Access Management, and confirm someone on your side holds Administrator, not just the agency. Do the same in Search Console under Settings, Users and permissions, where you want Owner-level access. Then export your last 16 months of Search Console query data now, while the relationship is still friendly. From what I have seen, the teams that skip this step end up negotiating blind, because every number they would cite in the conversation belongs to the party sitting on the other side of the table.
Where I think the model breaks
A couple of things nag at me. Kaushik's outcome-incentive tier, 15% to 25% of fees tied to organic revenue or verified lift, assumes you can measure incremental lift cleanly. Most attribution setups can't, and it gets worse now that a chunk of "organic" influence happens inside an AI answer nobody clicks through. Tie a bonus to a number you can't defend and you have basically built next year's argument in advance.
And to be fair, this isn't entirely new either. Performance-based agency pricing has been pitched for a decade and mostly fizzled, because both sides eventually game whatever metric gets picked. What feels different this time is that the base-retainer shrink is real and defensible, even while the incentive tier is still mostly aspirational. If it were my budget, I would take the guaranteed cut on the commoditized work and stay skeptical of the pay-for-lift promise until the measurement actually catches up. My guess: within 12 months, "AI citation share" shows up as a line item on more SEO retainers than "keyword rankings" does. The agencies that price that work now will set the rate everyone else pays later.
What I would bring to the next renewal
If your SEO renewal is coming up, don't open with a percentage. Open with a line-item split instead: a base retainer for governance and data, project fees for the entity and GEO work, and only then a small outcome slice that both sides actually trust. Kaushik's numbers (40-50% base, 30-40% project, 15-25% outcome) are a starting map, not scripture. The thing that gives the whole conversation teeth isn't the discount you ask for. It is walking in already holding your own GA4, your own Search Console, and 16 months of query data your agency didn't get to hand-pick. Get that part done this month. The fee talk goes a lot easier once the data is already sitting in your accounts.
Notice Me Senpai Editorial
