AI Ad Spend Is Exploding — But Are New York Businesses Getting Real ROI?
U.S. advertising spend is forecast to reach $414.7 billion in 2026, up 5% from 2025. J.P. Morgan And a growing, dominant share of that spend is flowing through AI-automated campaign tools — Google Performance Max, Meta Advantage+, and a growing suite of machine-learning-driven systems that promise to do the hard work of targeting, bidding, and optimization so you don't have to.
The pitch from both platforms is essentially the same: give us your budget, tell us your goal, and our AI will find the right people at the right time at the right price. Stop worrying about the mechanics. Just feed the machine and watch the results come in.
For large advertisers with massive data sets, deep conversion histories, and teams of specialists managing the machines, this pitch is largely true. For the typical small to mid-size New York business — the professional services firm spending $5,000 a month on Google Ads, the B2B company running Meta campaigns to generate leads, the local brand trying to compete in one of the most expensive advertising markets in the country — the reality is considerably more complicated.
This post is not an argument against AI-powered advertising tools. Used correctly, they deliver real value. It is an honest, data-driven look at where they perform, where they fall short, and what New York businesses specifically need to understand before handing their entire ad budget to an algorithm.
The Platforms' Interests and Your Interests Are Not the Same Thing
Before we get into the mechanics of specific tools, it's worth naming the structural tension that sits at the center of every conversation about AI-automated advertising.
Google and Meta are not your marketing partners. They are businesses whose revenue comes from advertising spend. Every feature they build, every default setting they configure, every recommendation their systems surface is calibrated first toward their own business outcomes — maximizing the amount of money flowing through their platforms — and second toward your business outcomes.
This is not a cynical conspiracy theory. It's just business. And acknowledging it is the foundation of using these tools intelligently rather than naively.
Google's optimization suggestions prioritize Google's revenue, not necessarily yours. Evaluate recommendations against your business objectives before implementing. ALM Corp That's not a third-party skeptic making that observation — it's a documented finding from analysis of advertiser accounts across 2025. The recommendations that Google's own interface surfaces to advertisers are not neutral. They are designed to increase spend.
Amazon, Google, and Meta have automated targeting, creative optimization, and performance reporting to the point where advertisers with straightforward direct-response needs can often bypass agencies entirely. J.P. Morgan That efficiency is real. But the same automation that removes friction from the buying process also removes visibility. When an algorithm is making thousands of micro-decisions per day about where your money goes, what it buys, and who it reaches, you need to be intentional about what oversight you maintain — because the platform's defaults are not designed with your specific business profitability in mind.
Performance Max: What the Data Actually Shows
Google Performance Max is now the dominant campaign type on the platform. By early 2025, over 73% of advertisers were running at least one Performance Max campaign. Google launched PMax in late 2021 and forcibly migrated Smart Shopping and Local campaigns into PMax in 2022 — making it nearly impossible to avoid. Savo Group
The premise of Performance Max is genuinely compelling: one campaign that runs across Google's entire inventory — Search, Display, YouTube, Discover, Gmail, Maps — and lets Google's AI decide in real time where to allocate your budget for maximum conversions. No more managing separate campaigns for each channel. No more manual bid adjustments. Just set your goal and let the machine optimize.
Here is what independent research actually shows about how that works in practice.
For e-commerce with strong data, PMax performs well. The Smarter Ecommerce study of 4,000-plus campaigns found that most PMax campaigns achieve between 95% and 116% of their ROAS targets. Savo Group For retailers with clean product feeds, high conversion volume, and strong historical data, Performance Max is a genuinely powerful tool that surfaces inventory across channels that manual campaigns couldn't efficiently manage.
For lead generation — which is most New York B2B businesses — the picture is significantly worse. The Adalysis study of 3,300-plus non-retail Performance Max campaigns found that when PMax and Search competed for the same terms, Search had higher click-through rates 65% of the time, higher conversion rates 84% of the time, and higher conversion values 84% of the time. PMax generated more raw impressions, but impressions without conversions do not pay the bills. Savo Group
That finding deserves emphasis. For the majority of New York businesses using Google Ads to generate leads — not e-commerce sales — Performance Max is likely underperforming compared to well-managed Search campaigns on the metrics that actually matter: qualified clicks and actual conversions.
The transparency problem is real and largely unresolved. Google added channel performance reporting in November 2025, which breaks down spend by channel. However, this is read-only data. You still cannot influence delivery by channel, adjust bids per specific placement, or prevent spend from flowing into low-converting placements. You can now see that 60% of your budget went to Display, but you cannot redirect it to Search. Transparency without control is a half-measure. Savo Group
The cannibalization problem is pervasive. Analysis of 503 accounts found that 91.45% had keyword overlap between Search and PMax campaigns. More than 56% of Search campaigns and 27.86% of ad groups were affected. Savo GroupIn plain language: PMax is frequently competing against your own Search campaigns for the same traffic, in some cases winning that competition and reporting the conversion as a PMax success — even though you would have gotten that click anyway through your Search campaign at a lower effective cost.
Small budgets and low conversion volume are severely penalized. Google and industry analyses recommend at least 30 conversions per month for stable PMax results, with 60 or more being optimal. Campaigns below 30 conversions show extreme variance. Xictron For a New York small business spending $3,000 to $5,000 per month on a service with a modest conversion rate, hitting 30 conversions per month may not be realistic — which means the algorithm is operating with insufficient data to optimize well, and the results will be inconsistent and unpredictable.
Meta Advantage+: The Real ROI Picture
Meta's Advantage+ suite follows similar logic: give the platform's AI control over audience targeting, creative selection, budget allocation, and bidding, and it will find conversions more efficiently than manual management.
The platform's own data on this is positive. Meta reports an average 9% lower cost per action with Advantage+ campaigns compared to manual setups. Coinis Independent analysis has supported this directionally for e-commerce. Meta conducted analysis of over ten thousand ad accounts using Advantage+ Shopping campaigns. The data revealed that automated setups delivered 32% more conversions compared to human-managed campaigns, with cost per acquisition dropping by 17% overall. ingeniom
But several important caveats significantly complicate this picture for New York businesses.
The data threshold requirement is the first and most important issue. Meta Advantage+ actually works at 50 conversions per week for most campaign types — or just 10 for purchase-optimized campaigns, a change most advertisers missed. Sandyriev A business generating 15 leads per week from Meta, with many of those leads not completing a purchase tracked by Meta's pixel, may be running Advantage+ with insufficient signal for the algorithm to optimize meaningfully. The AI is guessing when it should be learning.
Targeting control has been deliberately reduced. Meta has consolidated detailed targeting categories, merging specific interests into broader groups. This shift means advertisers have less granular control over audience targeting, which could affect campaigns aimed at niche audiences. Enrichlabs For a New York B2B company trying to reach, say, CFOs at professional services firms in Manhattan with 50 to 200 employees, the targeting precision that made Meta useful for sophisticated B2B campaigns has been meaningfully reduced. The platform is steering advertisers toward broader audiences that generate more impressions — and more platform revenue — while pushing the "trust the algorithm" narrative.
Attribution inflation is a documented problem. Meta's in-platform reporting consistently overstates the business impact of its campaigns compared to more neutral measurement approaches. A comprehensive audit of real-world accounts revealed that 30.67% of purchase conversion data is vanishing before it reaches Google's algorithms. Safari and Firefox already block 34.9% of tracking. Sandyriev This measurement degradation cuts both ways: you're losing signal on real conversions while Meta's attribution model may be claiming credit for conversions that were driven by other channels or would have happened organically. The ROI numbers in your Meta dashboard are almost certainly more optimistic than your actual business outcomes.
The creative dependency is real and often underestimated. Accounts with 4 images and 1 generic headline get destroyed by competitors with 20 diverse images, 6 videos, and 15 strategically crafted headlines. Google's algorithm needs options to test. If you provide limited assets, it has nothing to optimize. ALM Corp The same logic applies to Meta. AI-automated campaigns perform in proportion to the quality and diversity of the creative assets you feed them. Businesses that launch Advantage+ with two static images and one piece of copy, then wonder why results are poor, are essentially asking an algorithm to optimize with its hands tied behind its back.
The New York Market Makes This More Expensive Than Anywhere Else
Here's the dimension of this issue that is specific to New York businesses and rarely discussed in national advertising guides.
New York is one of the most expensive paid advertising markets in the country. CPCs for competitive keywords in finance, legal services, real estate, professional services, healthcare, and B2B technology — the categories that dominate New York's economy — are among the highest in any market globally.
When an AI-automated campaign is running with insufficient conversion data, or when it's allocating budget to Display and YouTube rather than high-intent Search, or when it's cannibalizing your organic traffic and attributing those conversions to itself, the cost of those inefficiencies is dramatically amplified by New York's already elevated ad costs.
A $3,000 monthly Google Ads budget in a mid-tier market might sustain enough conversion volume for PMax to optimize adequately. The same $3,000 budget in New York — where a single click on a competitive professional services keyword can cost $15 to $40 — buys far fewer clicks, which means far fewer conversions, which means the algorithm is perpetually data-starved and your results are perpetually inconsistent.
A 2026 survey found that 42% of small businesses struggle with AI-automated campaigns producing irrelevant traffic that wastes ad spend. Manual campaigns allow for precise exclusions that automated systems sometimes ignore entirely. ingeniom In New York, where each wasted click on an irrelevant query costs two to five times what it would in most other markets, that 10 to 15% budget waste from inadequate negative keyword management is not a rounding error. It's a material portion of a small business's advertising budget flowing to the platform with zero business return.
What Actually Works: The Hybrid Model
The answer to all of this is not to abandon AI-automated tools. The answer is to use them strategically, with appropriate human oversight, in the specific contexts where they genuinely outperform manual management.
Here is the framework that the data supports for New York small to mid-size businesses.
Use automation for what automation is actually good at. Smart bidding strategies — Target CPA, Target ROAS, Maximize Conversions — are genuinely superior to manual bidding for accounts with sufficient conversion data. These algorithms process real-time signals that no human manager can match at scale. The bidding layer of Google Ads is legitimately where AI adds unambiguous value. Surrender that fight gracefully and use the saved attention for the things that actually require human judgment.
Maintain manual control over audience and placement strategy. The targeting and placement decisions — which keywords to include, which audiences to build, which placements to exclude, which negative keywords to implement — are where your market knowledge creates genuine value that the algorithm doesn't have. Setting and forgetting automated campaigns is a documented failure mode. Performance Max, AI Max, and Demand Gen require active management: reviewing performance weekly, refreshing creative monthly, and adjusting strategy quarterly. ALM Corp
Build your first-party data infrastructure before scaling automation. From an architectural perspective, these platforms are no longer ad interfaces but data ingestion engines. Their output directly reflects the quality of signals you provide. Sandyriev For New York B2B businesses specifically, this means connecting your CRM to your ad platforms, implementing offline conversion tracking, and feeding the algorithm data about what actually converts into closed business — not just form completions that may or may not turn into clients. Without this, Google optimizes for form fills and Meta optimizes for clicks, and neither is optimizing for your actual business outcomes.
Set meaningful guardrails before turning the algorithm loose. Campaign-level budgets with firm daily limits. Negative keyword lists that exclude irrelevant queries before the campaign launches, not after you've burned three weeks of budget on them. Brand exclusions so the algorithm isn't cannibalizing your organic traffic. Search theme signals in Performance Max that bias the AI toward your actual market rather than wherever it finds the path of least resistance. Combining search themes and negative keywords gives you 30 to 40% more control over where PMax budget flows in the search network. Savo Group
Never trust in-platform reporting alone. Both Google and Meta attribution models claim credit aggressively and measure through their own lens. Allocate at least 10% of your budget to incrementality testing to validate whether your ads drive new revenue — using methods like A/B tests or geo-holdout experiments that isolate the true lift of a specific campaign. Sandyriev This sounds complex but it doesn't have to be. Simply comparing weeks when your campaigns are running versus dark periods, controlling for seasonality, gives you a meaningfully more honest picture of actual impact than your Google Ads dashboard does.
Feed the creative machine properly. Whether you're running PMax or Meta Advantage+, the algorithm's performance ceiling is determined by the quality and variety of creative assets you provide. This means investing in real creative development — multiple strong headlines, multiple image formats, real video even if it's simple, landing pages that actually convert rather than generic service pages. The AI cannot create leverage from bad creative. It can only find the best version of what you give it.
The Honest Bottom Line
AI-powered ad platforms are powerful tools that genuinely improve results when deployed correctly. They are also tools that, when deployed naively — or when trusted entirely without meaningful human oversight — systematically transfer budget from your business to the platform.
For New York businesses operating in expensive, competitive markets with constrained budgets and high cost-per-click environments, the stakes of getting this wrong are real. Handing Performance Max a $5,000 budget with no guardrails, no CRM integration, no negative keyword strategy, and no offline conversion tracking does not produce the same outcome as the case studies Google publishes. It produces wasted spend distributed across channels you didn't choose, optimized toward metrics that don't reflect your actual business, with attribution numbers that make the dashboard look better than the bank account does.
The businesses getting genuine, sustainable ROI from AI advertising tools in New York right now have one thing in common: they treat the AI as a powerful executor of a strategy that humans define — not as a strategy unto itself. They feed the machine properly, they guard the inputs carefully, they measure the outputs honestly, and they make adjustments based on real business outcomes rather than platform-reported metrics.
That combination — AI's processing power guided by human strategic judgment specific to your market — is what actually works. Everything else is paying Google and Meta to optimize for their goals with your money.
Running Google or Meta ads and not sure whether your AI-automated campaigns are actually working for your business — or just for the platform?
Ritner Digital audits paid advertising accounts for New York businesses to identify where AI automation is adding genuine value and where it's creating expensive inefficiency. We build the guardrails, data infrastructure, and measurement frameworks that make AI ad tools work for your bottom line instead of the platform's.
Get a paid advertising audit from Ritner Digital →
Sources: Dentsu U.S. Ad Spend Forecast 2026, Smarter Ecommerce State of Performance Max 2025 (4,000+ campaigns), Adalysis Performance Max vs Search Study (3,300+ campaigns), Optmyzr Account Analysis (503 accounts, February 2025), Meta Advantage+ Analysis (10,000+ accounts), Search Engine Land Small Business AI Advertising Survey 2026, JP Morgan Advertising Agency Competitive Analysis 2026, WordStream-LocaliQ 2026 Benchmarks Report.
Frequently Asked Questions
Google keeps recommending I switch everything to Performance Max. Should I follow their recommendations?
This is probably the most important question any New York advertiser can ask right now, and the honest answer is: evaluate every Google recommendation with significant skepticism before implementing it. Google's in-platform recommendations are generated by systems that are optimized to increase platform revenue — not to maximize your business profitability. That doesn't mean every recommendation is wrong. Some are genuinely useful. But the pattern of recommendations consistently pushes advertisers toward higher spend, broader targeting, and less manual control — all of which benefit Google's revenue regardless of whether they benefit yours. The specific recommendation to consolidate everything into Performance Max is one that deserves particular scrutiny for lead-generation businesses. Independent research shows that Search campaigns outperform PMax on click-through rate, conversion rate, and conversion value the majority of the time for non-retail accounts. Before making any major structural change to your campaign setup based on a Google recommendation, ask yourself: does this change give me more or less visibility into where my money goes? Does it give me more or less control over who sees my ads? If the answer to both is less, proceed with caution and test in a limited way before committing.
Our Meta Advantage+ campaigns are showing great results in the dashboard. Why do you say attribution might be inflated?
Meta's attribution model is designed to claim credit for conversions as broadly as possible — which is in Meta's interest because it makes campaigns look more effective than they are, justifying continued and increased spend. Several specific mechanisms cause in-platform metrics to overstate real business impact. First, Meta's default attribution window is seven-day click and one-day view, meaning if someone saw your ad and then converted up to seven days later through any channel — organic search, direct visit, a referral — Meta counts that as their conversion. Second, significant tracking data is lost due to iOS privacy changes, Safari and Firefox cookie restrictions, and ad blockers, meaning Meta is reconstructing a portion of its reported conversions through statistical modeling rather than direct measurement. Third, Meta campaigns often reach people who were already likely to convert — existing customers, warm audiences, people who had already searched for your service — and then claim credit for conversions that would have happened anyway. The way to get an honest read on your Meta ROI is to run a true holdout test: pause Meta campaigns for a defined period in a specific geographic area or audience segment and measure whether business outcomes actually change. Most businesses that do this for the first time discover their Meta attribution is anywhere from 30% to 60% more optimistic than the real incremental impact.
What's the minimum monthly budget a New York business needs for AI-automated campaigns to work properly?
There's no universal number, but there are data-based thresholds worth understanding. For Google Performance Max, the algorithm needs at least 30 conversions per month to optimize with any consistency — 60 or more being the point where performance genuinely stabilizes. In New York's expensive market, generating 30 monthly conversions from paid search requires enough budget to buy enough clicks at New York CPCs to produce that volume. Depending on your industry and what counts as a conversion, this might require anywhere from $5,000 to $15,000 per month or more. Below that threshold, PMax operates with insufficient data and produces highly variable, often disappointing results — not because PMax is bad, but because the algorithm is effectively guessing rather than optimizing. For Meta Advantage+, the platform needs 50 conversion events per week for most campaign objectives to optimize meaningfully. A New York B2B company generating 8 to 12 leads per week from Meta almost certainly isn't hitting that threshold, which means Advantage+ is running with inadequate signal. The practical implication is that smaller budgets often perform better under more controlled, manually-managed campaign structures where every dollar is deliberately directed at high-intent audiences rather than left to an algorithm that lacks the data to make good decisions.
We've been running Performance Max for six months and can't tell where our budget is actually going. Is that normal?
Unfortunately, yes — and it's one of the most documented and legitimate criticisms of PMax. Google added channel-level reporting in late 2025, which lets you see how spend was distributed across Search, Display, YouTube, Discover, Gmail, and Maps after the fact. But that reporting is read-only. You can see that 55% of your budget went to Display and YouTube last month, but you cannot set a rule preventing that from happening next month. You cannot increase the allocation to Search where your conversions are actually coming from. You cannot exclude specific Display placements that are burning budget with zero conversions. This transparency-without-control architecture is a deliberate design choice that keeps advertisers dependent on Google's optimization decisions rather than their own. The partial workarounds that have emerged — search themes to bias the algorithm toward your actual market, campaign-level negative keywords, brand exclusions to prevent cannibalization of organic traffic, asset group-level signals — help at the margins but don't solve the fundamental visibility problem. If you're running PMax alongside Search campaigns, the most important oversight step is checking for keyword overlap regularly: confirm that PMax isn't cannibalizing your Search campaign traffic and then claiming those conversions as its own wins.
Is it worth running both Google Ads and Meta Ads, or should we pick one?
For most New York businesses, the platforms serve genuinely different functions and the right answer is usually both — but with clear strategic logic for each rather than running both because everyone says you should. Google Search captures existing demand: people who are already looking for what you offer and have typed a relevant query. The intent signal is explicit. This is where you capture buyers who are already in the market. Meta creates and re-engages demand: it reaches people based on behavioral and demographic signals who may not be actively searching yet, either warming them toward your brand or retargeting people who have already expressed interest through website visits or content engagement. The sequencing that tends to work well is using Meta for top-of-funnel awareness and retargeting to people who engaged with your website, then using Google Search to capture the high-intent queries those same buyers eventually make after they've been exposed to your brand. The two channels reinforce each other when structured this way. Where the strategy breaks down is running both platforms in isolation — separate audiences, separate measurement, no connection between what someone saw on Meta and what they searched on Google — and then trying to evaluate each independently. The combined effect of well-coordinated cross-channel campaigns is greater than either in isolation, but only when the strategy is coherent across both.
Our industry has high CPCs in New York — finance, legal, real estate. How do we make AI advertising work when every click is expensive?
High-CPC environments actually demand more human oversight of AI tools, not less — because the cost of algorithmic inefficiency is amplified by every expensive click the system wastes. In practical terms, this means several things. First, negative keyword management is non-negotiable and must be treated as an ongoing maintenance task, not a setup-and-forget configuration. In high-CPC industries, a single wasted click on an irrelevant query can cost $20 to $50 or more. Even a 10% waste rate on a $10,000 monthly budget is $1,000 per month flowing to the platform with zero business return. Second, match type discipline matters more in expensive markets. Broad match keywords fed into Performance Max in a high-CPC industry will generate expensive, irrelevant traffic faster than in a low-CPC environment. Phrase and exact match in Search campaigns, combined with PMax search themes that are tightly defined, give you meaningfully more control over what triggers your ads. Third, landing page quality is the highest-ROI investment in expensive ad markets. When clicks cost $30, a landing page that converts at 8% rather than 4% doesn't just improve your results — it effectively cuts your cost per lead in half, making your entire campaign economics viable where they previously weren't. No amount of AI optimization compensates for sending expensive traffic to a generic service page that doesn't convert.
Google says Performance Max drove $20 billion in incremental revenue for advertisers in 2025. Doesn't that prove it works?
That statistic deserves careful scrutiny before it changes your strategy. "Incremental revenue" in Google's framing is measured using Google's own attribution methodology — which, as discussed, is designed to credit Google's platforms as broadly as possible. It aggregates across all PMax advertisers globally, including large e-commerce retailers with enormous conversion volumes and mature product feeds who are genuinely the ideal PMax customer. The average masks enormous variance: a retailer with $500,000 monthly budget, 10,000 monthly conversions, and a perfectly optimized product feed is having a very different PMax experience than a New York professional services firm with $8,000 monthly budget and 25 monthly lead form completions. Platform-level aggregate statistics also capture the performance of accounts managed by Google's own specialists and premium agency partners who have access to beta features, dedicated support, and optimization guidance that most small businesses never receive. When evaluating whether PMax is working for your specific business, the only number that matters is your own: is your cost per qualified lead going up or down, are those leads converting to clients at the rate your business needs, and is the revenue those clients generate covering your ad spend with the margin you require? Those numbers, measured honestly against what you'd be generating without the campaign, are your ROI — not Google's aggregate.
What's the single most important thing we can do to improve our AI ad campaign performance right now?
Connect your CRM data to your ad platforms and implement offline conversion tracking. This is the answer for the majority of New York B2B businesses, and it is underimplemented to a striking degree. Here's why it matters so much: when you run Google Ads or Meta campaigns without offline conversion tracking, the platform's AI is optimizing toward whatever in-platform event you've defined as a conversion — usually a form submission or a phone call click. But the business outcome you actually care about is closed revenue: the form submissions that turned into discovery calls, the discovery calls that turned into proposals, the proposals that turned into clients. Those downstream events are invisible to the algorithm unless you feed them back in. Without that data, Google optimizes for form fills regardless of quality. You end up with campaigns that are very efficient at generating form completions from people who are not actually qualified buyers — because that's what you told the algorithm to find. With offline conversion tracking connected to your CRM, you can feed back data on which leads became clients and at what revenue value, and the algorithm shifts toward finding more people like your actual customers rather than your form-fill volume. This single change — properly implemented — typically produces more meaningful campaign improvement than any targeting adjustment, budget increase, or campaign structure change you can make.