Autopilot Advanced · Learning Protection · Scheduled Tasks · Best Practices
Learning-period protection, "watch list" categories, TikTok product coverage analysis, scheduled tasks, 5 best practices + 3 things to avoid + FAQ.
New ads get "protected" and left untouched
The system protects ads still in their learning period from being disturbed too early, and cross-validates across multiple dimensions before suggesting any change. Specific rules may evolve as the product iterates.
- Ads still in the learning period, AI won't touch at all — no budget increases, no pausing, nothing. Let them run long enough to produce a baseline first.
- Meta gets an extra layer of insurance: as long as Meta's back end still shows the ad as "learning," AI skips it too — respecting the platform's pace.
So you won't see these ads in the main report — but open the "Full execution log" at the bottom of the report and they're right there (marked "Skip reason: learning-period protection"). If you ever notice "AI didn't touch the ad I just created" — this is why.
What kinds of situations land in the "watch list"
Not every signal means you have to act right away — some just deserve a little attention. AI puts these in a separate "watch list," with a label on each row telling you why it flagged it:
| Label | In plain words | Which platforms |
|---|---|---|
| Underspending | Budget isn't getting used up — AI reminds you to check whether the ROAS target is set too high, or whether the product itself just isn't competitive | Meta + TikTok |
| Near scaling threshold | Close to the "ready to scale up" bar but not quite there — AI suggests watching it a few more days | Meta + TikTok |
| Recent slowdown | Short-term performance is weaker than the mid-term baseline — AI holds off on suggesting a budget increase for now | Meta + TikTok |
| Recent deterioration | Short-term numbers look rough but mid-term is still OK — AI won't directly suggest pausing, it leaves the call to you | Meta |
| auto_budget taking over | It should be scaling up, but TikTok's own auto-scaling feature is already on it, so AI doesn't intervene twice | TikTok |
| auto_budget at threshold | TikTok's auto-scaling is getting close to its ceiling — you may need to adjust manually | TikTok |
TikTok exclusive: the "missing products" analysis
In GMV Max mode AI scans every product (SPU) on sale in your store, compares them against the ads you currently have running, and mainly looks at two things:
A · Which products are "running bare" (on sale but with no ads)
A product is on sale in your store, but no GMV Max ad is promoting it — AI suggests building an ad for it.
B · Which products earn especially well inside an ad (worth breaking out to run on their own)
An ad has several products in it, and one of them has a standout ROAS — AI suggests breaking it out into its own ad with its own budget, so it can run more steadily.
TikTok platform rule: the same product can only be covered by one GMV Max ad at a time — so you'll never end up with "two ads running the same product at once."
Meta and TikTok are managed separately
Autopilot treats the two platforms as completely independent:
- It won't proactively suggest "Meta's doing well, move TikTok's budget over" — that kind of cross-platform reallocation isn't something it does
- It won't merge the two sides' data into a combined analysis
- Checks have to be triggered separately too — "Run a check on my Meta" and "Run a check on my TikTok" are two separate flows
Want a cross-platform comparison? It's not within Autopilot's scope, but you can ask AI for a one-off comprehensive analysis on its own — "Compare how my Meta and TikTok are doing and give me a budget allocation suggestion" — that's a strategy-level diagnosis.
What about Google Ads
Google Ads AI-inspection auto-optimization isn't available yet, but you can still change status, budget, and bid targets through manual operations — and for automated rule execution, use Custom Rules (pause / budget / bid / notify all supported). When you need to make batch adjustments, just tell GrowthGPT what you want directly in the chat.
Scheduled tasks · turn ads on and off in plain language
Beyond data-driven optimization, Autopilot also supports scheduled ad on/off. Just describe your timing needs in plain language; it supports both one-time and daily / weekly recurring tasks:
| Scenario | Example command |
|---|---|
| Auto-stop when a promo ends | "Pause this campaign at 11pm on May 15" |
| Daily 9am–11pm | "Auto-activate at 9am every morning, auto-pause at 11pm every night" |
| Save budget on weekends | "Pause early Saturday morning, activate Monday morning" |
✅ 5 little habits that make this smoother
- Run a check every 1-2 days — don't wait until something goes wrong to look back; build a regular habit
- Run a diagnosis before turning on Autopilot — send a line like "Analyze the overall performance of my ad account" to understand your baseline first
- Don't rush to let AI touch a new ad — AI skips ads in the learning period anyway; give the algorithm time to learn
- Check "Needs your action" items right away — these are things AI can't solve, and leaving them unhandled drags down your ads' progress
- Read the full Summary Report once a week — gradually get a feel for AI's judgment logic, and slowly trust it to do more
❌ 3 things you're better off not doing
- Don't edit the same ad in GrowthGPT and the ad platform's back end at the same time — changing it on both sides at once is asking for conflicts
- Don't ask AI to optimize an ad the moment it goes live — there isn't enough data yet, so its judgment will be off
- Don't ignore "Needs your action" reminders — issues AI can't solve only pile up higher the longer you leave them
FAQ
Q1: Will Autopilot spend my money on its own?
No. Every budget change and ad pause requires your explicit confirmation before it runs. GrowthGPT only analyzes and recommends — the final decision stays in your hands.
Q2: How much does it scale up by?
The system calculates a reasonable amount based on your ad performance data, using a gradual scaling strategy — never a big jump all at once. Every suggestion clearly shows the before-and-after amounts.
Q3: What metrics decide whether an ad should be paused?
The system cross-validates across multiple dimensions before suggesting a change — weighing spend and conversion performance over both the short-term and the mid-term, combined with your target CPA / ROAS. If it's just short-term volatility while the mid-term still holds up, it'll be tagged "recent deterioration · monitor it" rather than paused outright. Specific rules may evolve as the product iterates.
Q4: Why didn't AI touch the ad I just created?
To protect the ad platform's learning period, campaigns still in the learning period won't have any operations performed on them. This is deliberate — AI only evaluates after the algorithm has accumulated enough data. To check the status, expand the "Full execution log" at the end of the report.
Q5: Can I use only some of the features?
Yes. You can have AI do diagnosis only, without executing anything; you can also accept budget-increase suggestions while rejecting pause suggestions — it's entirely up to you.
Q6: Does it support multiple ad accounts?
Yes. If you've connected multiple accounts, Autopilot scans and analyzes each account separately, and the report is broken out by account too.
Remember: Autopilot isn't "self-driving," it's "cruise control + auto-follow" — you're always in the driver's seat, and AI just cuts down on the tiring repetitive work.
What Autopilot Is · 3 Steps to Get Started
Set goals and guardrails — let AI inspect your ad account every day, surface opportunities, cut risk losses, and generate optimization suggestions you can execute with one click.
Custom Rules Engine — Complete Guide
Autopilot's rule-based execution capability: define IF…THEN… conditions and actions, and let rules run automatically on schedule — from creation to activation to reviewing results, all in one guide.