Quick answer: what should you ask before buying an AI ad optimization tool?
Before buying an AI ad optimization tool, ask what it does that Advantage+ and Performance Max no longer leave to you. Then ask how its results were measured after Meta's 2026 attribution changes and whether it will run a holdout judged on your own delivered revenue. Finally, check what it can change in your account, the full cost and the exit terms.
Why the buying questions changed in 2026
The platforms now do much of what tools used to sell, and Meta has reportedly changed how conversions are counted. A checklist written before 2026 misses the 2 things most likely to mislead you.
- Platform AI is the default. In early February 2025 Meta announced a test of a single setup for Sales, App and Leads campaigns with Advantage+ on by default. It also renamed Advantage+ shopping campaigns to Advantage+ sales campaigns.
- Performance Max (PMax) gained controls. Campaign-level negative keywords rolled out from January 2025, with the limit raised from 100 to 10,000 per campaign in March 2025. Trade press reports that Google Ads Editor 2.11 added PMax search term reports. We could not confirm this from Google.
- Attribution changed twice. Trade coverage citing Meta's developer changelog reports that Meta removed the 7-day and 28-day view windows from its reporting API on 12 January 2026. Trade coverage quoting Meta's business blog reports that from 3 March 2026, click-through counts link clicks only.
For what these tools are and what they typically cost, start with our complete guide to AI ad optimization tools. This article is the checklist you take into the vendor call.
Questions 1-3: make the vendor prove what it adds
These 3 questions decide whether there is anything to buy. If the vendor cannot answer them clearly, the rest of the call does not matter.
Q1. What does your tool do that Advantage+ or Performance Max doesn't already do?
Ask for specific levers, not adjectives. Good answers include moving budget between Meta and Google on one profit view, structured creative testing, profit-based pause and scale rules, and product feed fixes.
A warning sign is "AI audience targeting" that simply repeats Advantage+ audience. Our comparison of paid tools with Advantage+ and Smart Bidding lists what the platforms already do for free.
Q2. Which attribution settings are your case studies measured on?
A before-and-after ROAS (return on ad spend) chart that spans January or March 2026 compares two different definitions. Part of any drop or gain may come from Meta's counting change, not the tool.
AdRise Lab reports that since 3 March 2026 the default for website-conversion campaigns is 7-day click, 1-day engage-through, 1-day view. We could not confirm this default from Meta. Ask for results on 7-day click only, so both periods use the same rule.
Trade press reports that Meta's API returns empty data, not an error, when a tool requests a removed view window (source). Ask whether any number in the case study came from such a field.
Q3. Will you run a holdout against a control campaign, judged on my backend revenue?
A holdout splits comparable budget between campaigns the tool manages and a control it does not touch. Judge it on Shopify or CRM (customer relationship management) revenue net of cancellations and returns, never on the vendor's dashboard.
In India, the Gulf and parts of South-East Asia, count delivered COD (cash on delivery) orders only. Orders returned unpaid, called RTO (return to origin), cost you shipping both ways.
- Agree the success metric in writing before signing: delivered net revenue or contribution margin.
- Split budget 50/50 if you can, and never give the control less than 30%.
- Run at least 4 weeks, avoiding Diwali, Ramadan or BFCM (Black Friday and Cyber Monday), when CPMs (cost per 1,000 impressions) jump.
- Freeze creative launches across both sides, or launch the same creatives in both.
- Close the books after your normal return and COD delivery window has passed.
Our rule of thumb: the tool passes only if its extra contribution margin beats its full monthly cost in both halves of the test, not just on average.
Questions 4-6: check control, profit logic and data plumbing
These questions decide how much risk you take on after signing. Most expensive mistakes come from write access, the wrong goal or broken data.
Q4. What can the tool change in my account, and what stops it?
Ask whether access is read-only or write. If it is write, ask about daily spend caps, approval flows, a full change log and one-click rollback.
Then ask which safeguards Meta or Google enforce and which exist only in the vendor's code. This matters more as AI agents get direct account access. Some vendors say agents can connect through a Meta MCP (Model Context Protocol) server, but we could not confirm this from Meta. Ask the vendor to show you the platform documentation, not a slide.
Q5. What does it optimize for: platform ROAS or my profit?
A tool that chases platform ROAS can scale campaigns that lose money. Ask whether it uses break-even ROAS after product cost, shipping, tax and RTO, and whether it values new and returning customers differently.
Also ask whether it sends value signals back to Meta and Google through server-side tracking, or only reads their reports. This is how we built ScaleWin: every ad is judged against break-even ROAS, not the platform's number.
Q6. How does your data connection handle platform API changes?
Platform changes break dashboards quietly. Ask the vendor how it handled each of these:
- The reportedly removed Meta view windows, and the history limits trade press reports Meta added in January 2026 (6 months for frequency breakdowns, 13 months for unique counts).
- The retirement of the old Advantage+ shopping APIs. Trade press reports that after Marketing API v25.0, shopping ad sets with more than 50 ads cannot be duplicated through the API, and that the per-ad-set limit fell from 150 to 50. We could not confirm this from Meta.
- Google Ads API version upgrades, and PMax features such as campaign-level negatives that arrive through them.
A good vendor names the person who watches deprecations and shows you a past customer notice.
Questions 7-10: fit, full cost, exit terms and accountability
The last 4 questions decide whether the tool fits your business and what it really costs to run and leave.
Q7. Does it fit my channel mix and account setup?
If you run Meta and Google, ask whether the tool treats them as one P&L (profit and loss) or as two silos. Cross-border brands and agencies should check multiple accounts, currencies and markets in one view.
Q8. What is the total cost of ownership?
Ask whether pricing is flat or a percentage of spend, and where the tiers jump. Then add ad tax: 18% GST in India, 10% GST in Australia, GST/HST in Canada by province, and VAT in the UK, the EU and the Gulf states that levy it, depending on country and tax ID. US buyers should check how sales tax applies to their invoices.
In India, Google Ads Help explains that invoices from Google's Indian entity show 18% GST, while invoices from a foreign entity such as Google Asia Pacific can show no GST when a valid GSTIN is on file, and you then self-assess 18% IGST under reverse charge. Advertisers report a similar split between Meta's Indian entity and Meta Platforms Ireland, but we could not confirm this from Meta, so check your own invoices. Either way, input tax credit needs a GSTIN (Google Ads Help).
Worked example: the true monthly cost for a ₹30 lakh account
Step 1: Price the fee at today's spend and the next tier
Example: an Indian D2C brand spends ₹30 lakh a month (₹1,00,000 a day) across Meta and Google, with an average order of ₹1,500. A vendor quotes 3% of spend, so ₹90,000 today. At ₹50 lakh of spend, that becomes ₹1,50,000.
Step 2: Add team time and onboarding
Weekly reviews take 8 hours a month at ₹2,000 an hour, so ₹16,000. Onboarding takes 30 hours, or ₹60,000, spread over 12 months: ₹5,000. Total monthly cost is ₹1,11,000, plus FX charges if the vendor bills in another currency.
Step 3: Work out the lift the tool must create
Assume a contribution margin of 35% on delivered net revenue, after product cost, shipping, payment fees and RTO.
Required extra delivered revenue
Monthly tool cost ÷ contribution margin = ₹1,11,000 ÷ 0.35 ≈ ₹3,17,000 a month, about 211 extra delivered orders at ₹1,500Step 4: Check the cash flow on scaling
If the tool recommends adding ₹10 lakh of spend, cash out rises by ₹11,80,000 with 18% GST. The GST is recoverable as input tax credit, but only later and only with a GSTIN.
Q9. Who owns the data, audiences, rules and creative, and what happens when I leave?
Get it in the contract. Is your data used to train models? How long is it kept, and is it deleted when you leave? Can you export rules, audiences, creative and change history in a usable format? What is the notice period?
Q10. Who on your side is accountable, and who on mine?
Ask for a named contact, support hours in your time zone (IST, Gulf, GMT, ET or PT, AEST, SGT) and an escalation path. Agree how often someone checks the tool's output and which decisions stay human, such as offers, pricing and festive budgets.
What most guides get wrong about buying an AI ad tool
Most ranking checklists are generic AI-procurement lists. They are useful for data readiness but miss the ad-specific traps.
- They ignore platform overlap. Few ask what the tool adds beyond Advantage+ and PMax, which is now the first question.
- They use old names and limits. Some still describe Advantage+ Shopping as a separate campaign type, date the unified setup to 2026 instead of February 2025, or say PMax has no negatives or caps them at 100. All of that is out of date.
- They treat vendor numbers as research. Figures such as "keep tool cost at 5-10% of spend" or "expect a 15-40% reported conversion drop" come from vendor blogs. Use your own Step 3 maths instead.
- They call the view-window removal sudden. Trade coverage reports that Meta flagged it in its developer changelog ahead of time. A vendor that never mentioned it was not watching.
- They trust platform ROAS. In COD-heavy markets, placed orders overstate revenue. A tool judged on them can look better than it is.
For the rules a good tool should use once you buy it, see how a tool decides what to pause, scale or fix.
Scorecard: good answers versus warning signs
Take this table into every vendor call. Score each answer as good or a warning sign, then apply the knock-out rule below.
| Question | Good answer | Warning sign |
|---|---|---|
| 1. Beyond platform AI | Names levers Advantage+ and PMax don't touch | "Smarter audiences" only |
| 2. Attribution used | 7-day click, same rule before and after | Chart spans 2026 changes, no setting stated |
| 3. Holdout | Yes, on your delivered net revenue | Only a dashboard trial |
| 4. Write access | Caps, approvals, change log, rollback | "The AI handles it" |
| 5. Goal | Break-even ROAS, margin, RTO | Platform ROAS only |
| 6. API changes | Named owner, past customer notices | Unaware of 2026 deprecations |
| 7. Fit | Meta and Google on one P&L | Single channel, single currency |
| 8. Total cost | Clear tiers, written quote | Fee jumps hidden in tiers |
| 9. Exit | Full export, deletion in contract | Audiences and rules stay with vendor |
| 10. Accountability | Named contact, your time zone | Ticket queue only |
A warning sign on question 2, 3 or 4 ends the evaluation. Otherwise, buy only if the vendor scores good on at least 8 of 10 and the holdout beats the Step 3 lift.
Read next in this series
- AI Ad Optimization Tool for Agencies: How to Manage Many Client Accounts From One Dashboard
- AI Ad Optimization Tools: The Complete 2026 Guide (What They Do, What They Cost, Who Needs One)
- AI Ad Optimization Tool vs Meta Advantage+ and Google Smart Bidding: What the Platforms Already Do for Free
- How an AI Ad Optimization Tool Decides What to Pause, Scale or Fix (The Rules Explained)
- Net Profit Calculator - Free net profit calculator for D2C and Shopify brands
Key takeaways
- The first question is what the tool adds on top of Advantage+ and Performance Max, which now handle audiences, placements and budgets by default.
- Trade press reports that Meta changed attribution on 12 January and 3 March 2026, so any vendor case study that spans those dates may mix two definitions of a conversion. Ask which attribution setting it used.
- Insist on a holdout against a control campaign, judged on backend revenue net of returns and undelivered COD orders.
- Work out the total cost: the fee at your next spend tier, team hours, ad tax cash flow and the cost of leaving.
- Questions 2, 3 and 4 are knock-outs. A vendor that fails any of them is out, however good the demo looks.
Ai ad optimization tool: frequently asked questions
Do I need an AI ad optimization tool if I already use Advantage+ and Performance Max?
How much should an AI ad optimization tool cost relative to ad spend?
How long should a trial or pilot run before judging an AI ad tool?
Why did my Meta conversions change in 2026, and does it affect how I judge a tool?
Is it safe to give an AI tool write access to my ad account?
Are AI ad optimization tools worth it for COD-heavy D2C brands?
Sources
- Meta restricts attribution windows and data retention in Ads Insights API (PPC Land)
- Meta attribution: engage-through and click-through changes in 2026 (AdRise Lab)
- Meta Advantage+ sales campaigns (Meta for Business)
- About taxes for Google Ads in India (Google Ads Help)
Let ScaleWin read your ad data every day
ScaleWin checks your Meta and Google Ads daily, judges every ad on profit and tells you exactly what to change.
Start free, no card



