What will happen when Google (most likely) has to deploy billions in ad credits to advertisers who are participating in the Mass Arbitration for ten years of monopolistic excess profit?
As Marketing Agency Didit Founder Kevin evaluated for clients he’s submitted into the Keller Postman Mass Arbitration for Ad Spend Recovery, Kevin having built SEM bid management and ad automation systems in the past was very familiar with the auction dynamics of Paid Search and Programmatic Display. So, he ran simulations of just what would happen if billions of dollars in ad credits resulting in hundreds of billions in ad spend recovery claims by advertisers were to flow back into the ad auction markets.
As you might expect the introduction of non-cash ad credits into the ecosystem results in significant inflation for ALL advertisers competing against the keywords and targets of the advertisers deploying those credits. When those credits are deployed strategically over time, the elasticity of the ad auctions result in the majority of the ad credit value returning to Google via ad inflation, mitigating the downside to Google from paying back advertisers for previously charged monopolistic profit. Oddly this makes the auctions even more unfair to the advertisers who are NOT receiving credits. Their hard dollars are competing with ad credits.
Kevin ran simulation scenarios with Chat GPT, Claude, an even Google’s Gemini. While the simulations made slightly different assumptions with regard to price elasticity.
For those analysts and investors in Google looking to understand the implications, we’ve published the multi-layered analysis from ChatGPT below.
Dialog with ChatGPT Pro:
Prompt: Google is engaged in a mass arbitration for monopolistic spend recovery. It’s likely that the settlement will will be paid to the advertisers in ad credits for google paid search. Simulate the impact and revenue increases across all SEM google revenue as a result of 5 billion, 10 billion, 20 billion, 50 billion and 100 billion of settlement credits is deployed into the auctions resulting in escalation of CPCs across ALL advertisers resulting in significant spending of both the settlement credits and real dollars competing in the adrank powered paid placement auctions. Specifically model whether or not the net cost to google of the settlement will be partially of fully mitigated by escalation in revenue due to surges in spending.
First Answer:
I’m pulling current Google Search ads revenue context and setting up a simple auction-spillover model so I can quantify when ad-credit “settlement dollars” turn into incremental real-dollar revenue for Google.
I’ve anchored the baseline at $224.5B of 2025 Google Search & other revenue, with paid clicks up 6% and CPC up 7%; now I’m testing two separate issues: auction spillover versus accounting treatment.
Treating the remedy as a hypothetical ad-credit settlement, my answer is:
On a face-value basis, Google probably does not fully earn it back through auction inflation. In my base case, the extra real-dollar revenue recaptured is about 45% of face value at $5B-$10B, drifting down to 37% at $100B as the market saturates.
On an economic-cost basis, smaller programs can get close to neutral. Because ad credits cost Google less than cash, a $5B-$20B program can be about 90% offset in my base case, while $50B-$100B is still negative, though partly mitigated.
I anchored the model to Google’s closest public SEM proxy: Google Search & other, which was $224.5B in 2025. Google says that bucket includes search properties plus other owned-and-operated properties like Gmail, Maps, and Play, so the true search-only denominator is somewhat smaller and search-only auction effects would likely be a bit larger than the table below. Google also reported paid clicks up 6% and CPC up 7% in 2025, and explicitly says advertiser competition for keywords affects Search monetization.
Google’s own auction docs also matter here: actual CPC is driven by Ad Rank, thresholds, and competition, and in the simplified example Google gives, the winner pays just enough to beat the next advertiser or the threshold. Google also exposes Search lost IS (budget) and says advertisers can improve impression share by raising budget or bids, which is the mechanism that makes credits capable of crowding in additional real-dollar spend from everybody else.
For cost, I use a 20% variable-cost proxy, anchored to Alphabet’s 20.3% 2025 TAC rate. That is conservative relative to Google Services’ overall profitability, which was $139.4B of operating income on $342.7B of revenue in 2025.
Base-case simulation
Assumptions:
- credits are deployed over 12 months
- 90% of credits are actually used
- 60% of redeemed credits are truly incremental auction pressure; 40% mostly substitute for spend those advertisers would have made anyway
- each $1 of net-new credit demand pulls in $0.85 of extra cash from the rest of the market at small scale, with diminishing returns as the program gets huge
- 80% of the above-baseline shock shows up as higher CPC, 20% as more paid clicks
Core formulas:
- Used credits:
U = 0.90 × C - Net-new auction pressure:
D = 0.60 × U - Incremental cash revenue:
ΔR = 0.85 × D / (1 + D / 224.5) - CPC uplift:
≈ 0.80 × (D + ΔR) / 224.5 - Credit economic cost: substituted cash spend
+variable cost on net-new credit volume= 0.40 × U + 0.20 × 0.60 × U = 0.468 × C
| Settlement credits | Incremental cash revenue to Google | Avg. CPC uplift across all advertisers | Net cash revenue change vs baseline | Economic cost of credits | Economic cost offset | Residual economic cost |
|---|---|---|---|---|---|---|
| $5B | $2.3B | 1.8% | +$0.5B | $2.3B | 93% | $0.2B |
| $10B | $4.5B | 3.5% | +$0.9B | $4.7B | 92% | $0.4B |
| $20B | $8.8B | 7.0% | +$1.6B | $9.4B | 90% | $1.0B |
| $50B | $20.5B | 16.9% | +$2.5B | $23.4B | 84% | $3.7B |
| $100B | $37.0B | 32.4% | +$1.0B | $46.8B | 76% | $11.3B |
What that means
For $5B-$20B, the model says Google can get most of the economic pain back via higher auction-clearing prices and competitor defense spending. In this range, the program can be close to profit-neutral, even though it is not face-value neutral.
For $50B, the shock becomes systemically large. You still get a big revenue response, but saturation kicks in: more advertisers start conceding position, reducing coverage, or letting impression share slip instead of matching every CPC increase. The settlement is still meaningfully mitigated, just not neutral.
For $100B, the settlement is simply too large relative to the current Search & other revenue base to self-fund in a one-year deployment. Even with a very strong auction response, the model still leaves a material net cost.
Why full face-value mitigation is hard
The key reason is auction mechanics. Because Google’s actual CPC is effectively the amount needed to beat the next competitor or the threshold—not your full bid—not every credit dollar becomes extra Google cash. Credits only generate large system-wide revenue if they move the marginal price-setting bidders across lots of auctions.
In small-shock terms, face-value recapture is approximately:
utilization × incrementality × crowd-in
In my base case that is:
0.90 × 0.60 × 0.85 ≈ 46%
To get to 100% face-value mitigation, you would need a much stronger crowd-in response—roughly $1.85 of extra competitor cash for every $1 of net-new credit demand under the same utilization and incrementality assumptions. That is a very high bar for a broad market-wide program.
Sensitivity range
If credits mainly go to advertisers who were already going to spend anyway, the recapture is weak: about 13%-15% of face value and only about 20%-23% of economic cost in my low case.
If credits are concentrated in budget-constrained, highly competitive verticals and competitors aggressively defend share, the recapture can be much stronger: about 60%-74% of face value, and potentially more than 100% of economic cost in an aggressive case.
Bottom line
My modeled answer is:
- Face-value settlement cost: partially mitigated, not fully
- Economic/operating-profit cost:
- $5B-$20B: close to neutral in a reasonable case
- $50B: substantially mitigated but still negative
- $100B: still clearly negative unless deployment is slow or unusually concentrated in hypercompetitive categories
The single biggest swing factor is incrementality. If the credits mostly replace spend advertisers would have made anyway, Google does not claw much back. If they hit budget-constrained accounts and force broad auction defense, Google can recover a lot more.
Refined Prompt:
re-evaluate and add a 200 billion and 500 billion settlement but assume that Google requires no more than 25% of the individual advertiser settlement is deployed in any calendar year so that credits are deployed over 48 months. Illustrate the same economic costs and recapture.
Refined Model:
I’m re-running the same model with a hard 25% annual usage cap, so credits deploy over 48 months instead of one year. That should mainly reduce saturation and annual CPC pressure, with the biggest recapture changes in the $50B-$500B cases.
The four-year pacing mostly leaves nominal credit cost unchanged while cutting the annual auction shock by roughly 75%, so the big cases become far more absorbable. In my first-pass numbers, $200B gets to about 88% economic recapture, while $500B still remains negative and appears to be the main break point.
I re-ran the model on a 48-month deployment schedule, with each advertiser capped at using no more than 25% of its award per calendar year. I kept the same structural assumptions as before and recomputed the whole stack consistently rather than just scaling the old outputs. For the annual denominator, I’m still using Google Search & other revenue of $224.5B in 2025. Google says that bucket includes Google search properties plus other owned-and-operated properties like Gmail, Maps, and Play, so a pure search-only view would likely show somewhat larger percentage shocks than the table below. Alphabet also reported paid clicks up 6% and CPC up 7% in 2025, and says Search monetization is affected by advertiser competition for keywords.
Mechanically, the same auction logic still applies: Google Ads says actual CPC is driven by Ad Rank, thresholds, and competition, and that Search lost IS (budget) measures the share of time ads did not show because budget was insufficient. That is the channel through which credits can both substitute for some existing advertiser cash spend and also crowd in incremental cash from other advertisers defending share.
For cost, I kept the same 20% variable-cost proxy, anchored to Alphabet’s 20.3% TAC rate and the 2025 profitability of Google Services, which had $342.7B revenue and $139.4B operating income. These are nominal, undiscounted economics, not GAAP litigation accounting.
The formulas are the same, except the shock is spread across four years:
- total redeemed credits:
U = 0.90 × C - average annual redeemed credits:
U / 4 = 0.225 × C - average annual net-new auction pressure:
D_y = 0.60 × U / 4 = 0.135 × C - annual incremental cash revenue:
R_y = 0.85 × D_y / (1 + D_y / 224.5) - 48-month incremental cash revenue:
4 × R_y - average annual CPC uplift:
0.80 × (D_y + R_y) / 224.5 - 48-month economic cost: displaced cash spend
+variable cost on net-new credited clicks= 0.40 × U + 0.20 × 0.60 × U = 0.468 × C
Updated 48-month simulation
| Settlement credits | Avg annual credits used | Avg annual CPC uplift | 48-mo incremental cash revenue | Face-value recapture | 48-mo net cash revenue change vs baseline | 48-mo economic cost | Economic-cost offset | Residual economic cost |
|---|---|---|---|---|---|---|---|---|
| $5B | $1.1B | 0.4% | $2.3B | 45.8% | $0.5B | $2.3B | 97.8% | $0.1B |
| $10B | $2.2B | 0.9% | $4.6B | 45.6% | $1.0B | $4.7B | 97.5% | $0.1B |
| $20B | $4.5B | 1.8% | $9.1B | 45.4% | $1.9B | $9.4B | 96.9% | $0.3B |
| $50B | $11.2B | 4.4% | $22.3B | 44.6% | $4.3B | $23.4B | 95.2% | $1.1B |
| $100B | $22.5B | 8.7% | $43.3B | 43.3% | $7.3B | $46.8B | 92.5% | $3.5B |
| $200B | $45.0B | 16.9% | $81.9B | 41.0% | $9.9B | $93.6B | 87.5% | $11.7B |
| $500B | $112.5B | 39.8% | $176.4B | 35.3% | -$3.6B | $234.0B | 75.4% | $57.6B |
What changed versus the 12-month shock
The total nominal economic cost of the credits is basically unchanged under the same utilization and incrementality assumptions. What changes is the efficiency of recapture. By limiting annual usage, Google reduces the size of each year’s auction shock relative to the annual Search revenue base, so less of the spend surge is lost to saturation. In this re-run, that meaningfully improves the big cases.
The easiest way to see it is the annual CPC shock. Under a 48-month rollout, the average annual uplift is much lower: 8.7% at $100B, 16.9% at $200B, and 39.8% at $500B. So the annual cap does what you want it to do: it converts a one-time auction shock into a more persistent but less explosive four-year inflation effect.
Interpretation
At $5B-$20B, Google is still very close to economically whole. The settlement is not face-value neutral, but the residual economic cost is trivial in this model.
At $50B-$100B, the settlement is still substantially mitigated. The annual cap makes these look much more manageable than a one-year dump into the auction. At $100B, Google recaptures $43.3B of incremental cash revenue over 48 months and offsets about 92.5% of economic cost.
At $200B, the slower deployment matters a lot. In this model it remains cash-positive on a baseline-comparison basis over the full 48 months, with $9.9B of net cash revenue improvement, though it still leaves $11.7B of residual economic cost once you include the variable cost of serving credited clicks.
At $500B, even spreading usage over four years is not enough to make the settlement self-funding. Google still recaptures a very large $176.4B of incremental cash revenue, but that only offsets about 75.4% of economic cost, leaving a residual cost of about $57.6B. On a pure cash-revenue basis, it also turns slightly negative at -$3.6B versus baseline.
Bottom line
Under this 25%-per-year usage cap, my re-evaluated answer is:
- $5B-$20B: effectively near-neutral on economic cost
- $50B-$200B: strongly mitigated, but not fully face-value neutral
- $500B: still materially negative, even with four-year smoothing
The cap does not eliminate the burden, but it materially improves recapture at the large settlement sizes by reducing annual auction saturation. A useful next step would be a sensitivity grid on incrementality and competitive crowd-in, because those two assumptions drive most of the spread once you get to $200B-$500B.
I’m turning the 48-month capped-deployment model into a two-variable sensitivity grid, varying incrementality and competitive crowd-in while holding the other assumptions fixed so you can see where mitigation becomes near-neutral.
The grid needs to be split by settlement size because saturation changes the answer, so I’m laying it out as size-specific matrices and a compact break-even threshold table. One pattern already stands out: under the 48-month cap, incrementality is the main swing factor, while the crowd-in required for break-even rises sharply only at the largest settlements.
Below is the 48‑month sensitivity grid using the same structure as the prior run. I kept the annual saturation denominator at 2025 Google Search & other revenue of $224.5B. Alphabet says that bucket includes search properties plus owned-and-operated properties like Gmail, Maps, and Play, says advertiser competition for keywords affects monetization, and Google Ads says actual CPC is determined by Ad Rank, thresholds, and competition.
I’m varying only these two levers:
- Incrementality (I): the share of redeemed credits that create net-new auction pressure rather than replacing spend advertisers would have made anyway.
- Competitive crowd-in (K): the amount of extra cash spend from other advertisers pulled in per $1 of annual net-new credit demand, before the saturation haircut.
Model equations:
E = U × [(1 − I) + 0.20 × I] economic cost to Google
D_y = U × I / 4 annual net-new credit pressure
R = 4 × [K × D_y / (1 + D_y / 224.5)] 48-mo cash recapture
Offset = R / E economic-cost offset
Residual cost = E − R
All figures below are nominal, undiscounted $B.
1) Economic cost by incrementality
Crowd-in does not change this table.
| Settlement | 30% | 45% | 60% | 75% | 90% |
|---|---|---|---|---|---|
| $5B | $3.4B | $2.9B | $2.3B | $1.8B | $1.3B |
| $10B | $6.8B | $5.8B | $4.7B | $3.6B | $2.5B |
| $20B | $13.7B | $11.5B | $9.4B | $7.2B | $5.0B |
| $50B | $34.2B | $28.8B | $23.4B | $18.0B | $12.6B |
| $100B | $68.4B | $57.6B | $46.8B | $36.0B | $25.2B |
| $200B | $136.8B | $115.2B | $93.6B | $72.0B | $50.4B |
| $500B | $342.0B | $288.0B | $234.0B | $180.0B | $126.0B |
Equivalent economic cost as a percent of face value is:
- 30% incrementality: 68.4%
- 45%: 57.6%
- 60%: 46.8%
- 75%: 36.0%
- 90%: 25.2%
So once you know the offset %, you can recover the other outputs directly:
Face-value recapture = Offset × (Economic cost / Settlement)
Residual economic cost = Economic cost × (1 − Offset)
2) Reference strip at base crowd-in = 0.85x
This is the same crowd-in assumption used in the prior run. Each cell shows:
48-mo cash recapture / face-value recapture
| Settlement | 30% | 45% | 60% | 75% | 90% |
|---|---|---|---|---|---|
| $5B | $1.1B / 23% | $1.7B / 34% | $2.3B / 46% | $2.9B / 57% | $3.4B / 69% |
| $10B | $2.3B / 23% | $3.4B / 34% | $4.6B / 46% | $5.7B / 57% | $6.8B / 68% |
| $20B | $4.6B / 23% | $6.8B / 34% | $9.1B / 45% | $11.3B / 57% | $13.5B / 68% |
| $50B | $11.3B / 23% | $16.8B / 34% | $22.3B / 45% | $27.6B / 55% | $32.9B / 66% |
| $100B | $22.3B / 22% | $32.9B / 33% | $43.3B / 43% | $53.4B / 53% | $63.2B / 63% |
| $200B | $43.3B / 22% | $63.2B / 32% | $81.9B / 41% | $99.8B / 50% | $116.7B / 58% |
| $500B | $99.8B / 20% | $140.5B / 28% | $176.5B / 35% | $208.5B / 42% | $237.3B / 47% |
3) Break-even crowd-in needed for 100% economic-cost offset
This is the cleanest “can Google be made whole?” view.
| Settlement | 30% | 45% | 60% | 75% | 90% |
|---|---|---|---|---|---|
| $5B | 2.54x | 1.43x | 0.87x | 0.54x | 0.31x |
| $10B | 2.54x | 1.43x | 0.87x | 0.54x | 0.31x |
| $20B | 2.55x | 1.44x | 0.88x | 0.54x | 0.32x |
| $50B | 2.57x | 1.45x | 0.89x | 0.55x | 0.33x |
| $100B | 2.61x | 1.49x | 0.92x | 0.57x | 0.34x |
| $200B | 2.69x | 1.55x | 0.97x | 0.61x | 0.37x |
| $500B | 2.91x | 1.74x | 1.13x | 0.73x | 0.45x |
Interpretation:
- At 60% incrementality, Google needs only about 0.87x–1.13x crowd-in to get economically whole across the full $5B-$500B range.
- At 45% incrementality, break-even gets much harder: about 1.43x at small scale and 1.74x at $500B.
- At 30% incrementality, even very aggressive competitor response is usually not enough.
4) Full 2D sensitivity grids: economic-cost offset %
Cells above 100% mean the settlement is more than fully mitigated on an economic-cost basis.
Base case from the prior run is the 60% incrementality / 0.85x crowd-in cell.
$100B settlement
| 0.50x | 0.85x | 1.20x | 1.50x | 2.00x | |
|---|---|---|---|---|---|
| 30% | 19 | 33 | 46 | 57 | 77 |
| 45% | 34 | 57 | 81 | 101 | 135 |
| 60% | 54 | 93 | 131 | 163 | 218 |
| 75% | 87 | 148 | 209 | 262 | 349 |
| 90% | 147 | 251 | 354 | 442 | 590 |
$200B settlement
| 0.50x | 0.85x | 1.20x | 1.50x | 2.00x | |
|---|---|---|---|---|---|
| 30% | 19 | 32 | 45 | 56 | 74 |
| 45% | 32 | 55 | 77 | 97 | 129 |
| 60% | 51 | 88 | 124 | 154 | 206 |
| 75% | 81 | 139 | 196 | 244 | 326 |
| 90% | 136 | 231 | 327 | 408 | 545 |
$500B settlement
| 0.50x | 0.85x | 1.20x | 1.50x | 2.00x | |
|---|---|---|---|---|---|
| 30% | 17 | 29 | 41 | 51 | 69 |
| 45% | 29 | 49 | 69 | 86 | 115 |
| 60% | 44 | 75 | 106 | 133 | 177 |
| 75% | 68 | 116 | 164 | 204 | 273 |
| 90% | 111 | 188 | 266 | 332 | 443 |
What this says
The model is telling a pretty clear story.
At the prior base case of 60% incrementality / 0.85x crowd-in, Google is:
- ~93% offset at $100B
- ~88% offset at $200B
- ~75% offset at $500B
That translates into residual economic cost of about:
- $3.5B at $100B
- $11.7B at $200B
- $57.5B at $500B
The most important lever is still incrementality.
If incrementality rises to 75%, even the base 0.85x crowd-in assumption fully offsets every settlement size up to $500B in this 48‑month deployment structure.
If incrementality is only 45%, then even pretty strong crowd-in is not enough at the largest sizes. At $500B, you need roughly 1.74x crowd-in just to break even economically.
If incrementality is only 30%, the model never gets Google whole anywhere in the 0.50x–2.00x crowd-in range, even with the 48‑month smoothing.
The practical takeaway is that the settlement becomes economically manageable only when a large share of credits are truly net new and/or the credits force a broad defensive cash response from the rest of the market. Without one of those two conditions, the dilution from substitution dominates.