An AI advertising optimization audit is a structured review of whether an ad account gives its automated systems accurate signals and sensible limits. Before changing bids or budgets, it checks conversion tracking, the values sent to platforms, account structure, audience inputs, creative supply and automation settings in Google Ads, Meta and similar tools. Most performance problems in automated accounts come from bad inputs, not from the algorithm itself.
Why audit before optimising?
Smart Bidding, Performance Max, Advantage+ and similar systems learn from the data you feed them. If a purchase fires twice, if leads of every quality count the same, or if the account mixes brand and non-brand traffic, the system optimises toward a distorted picture. Changing targets on top of that only moves the distortion around.
An audit also creates a baseline. Recording the state of the account before touching anything means you can tell later whether changes helped or whether results moved for outside reasons, such as seasonality.
Step 1: Is conversion tracking accurate?
Everything else depends on this. Start by listing every conversion action in each platform and checking what triggers it.
- In Google Ads, open Goals > Conversions > Summary. Check which actions are set as primary (used for bidding) and which are secondary. Page views or engaged sessions set as primary goals are a frequent problem.
- Compare conversion counts for the last 30 days with your source of truth: the store backend, CRM or booking system. Large gaps point to duplicates or missing events.
- Check that Enhanced Conversions (Google) and the Conversions API (Meta) are active and deduplicated with browser events.
- Confirm consent mode is implemented correctly in regions that require consent, so modelled conversions are available.
- Look at the tag diagnostics in each platform for errors or inactive tags.
Step 2: Are the conversion values right?
Value-based bidding, such as target ROAS or maximise conversion value, uses the values you send. If every lead is worth the same “1”, the system cannot tell a qualified enterprise lead from a spam form.
Check that purchase values match actual order values, including currency, and decide whether to send revenue or margin. For lead generation, consider offline conversion imports from the CRM so qualified leads or closed deals carry higher values. Google Ads supports this through offline conversion imports and the data manager; Meta supports it through the Conversions API with CRM events.
Step 3: Does the account structure help or hurt learning?
Automated bidding learns best when each campaign has enough conversions. Over-segmented accounts, with dozens of campaigns each getting a handful of conversions per month, keep the system in permanent learning.
Review these points:
- Number of conversions per campaign over the last 30 days compared with the platform’s guidance for the bidding strategy in use.
- Whether brand and non-brand search are separated. Brand traffic converts cheaply and can make non-brand performance look better than it is if mixed.
- Overlap between Performance Max and standard Shopping or Search campaigns.
- Campaigns with very different margins or goals combined under one target.
- Learning status and recent large changes listed in the change history.
How do you review audience and targeting signals?
Automated campaigns still use your inputs as signals. Audience signals in Performance Max, customer lists, website visitor lists and lookalike seeds all shape where the system starts looking.
Check that customer lists are recent and large enough to be used, that exclusions for existing customers are set where acquisition is the goal, that location settings use “presence” rather than “presence or interest” where that matters, and that negative keywords or brand exclusions are in place. In Meta, check whether Advantage+ audience is on and whether any old detailed targeting restrictions conflict with it.
Is there enough creative for the system to work with?
Automated systems rotate and combine assets. With too few, they have nothing to test, and performance falls as frequency rises.
- Responsive search ads: check ad strength and asset variety. Fifteen headlines that say the same thing in different words offer little real choice.
- Performance Max asset groups: review asset group strength and see which images, videos and headlines are rated low.
- Meta: count active creatives per ad set and check frequency and click-through trends over the last four weeks.
- Video: confirm there are vertical versions for Shorts, Reels and Stories placements.
- Landing pages: check that they load quickly on mobile, since slow pages lower conversion rates and the systems learn from that.
Automation settings and guardrails
Many automation features are on by default. Some help, others need a deliberate decision. Review auto-applied recommendations in Google Ads (Recommendations > Auto-apply), automatically created assets, final URL expansion in Performance Max, broad match settings and Meta’s Advantage+ creative enhancements.
None of these is wrong by default, but each should be a choice. Check that brand guidelines allow automatically generated text or image edits, and that URL expansion does not send traffic to pages such as careers or blog archives.
Document the baseline and plan changes
Finish the audit with a short written record: current spend, conversions, cost per conversion or ROAS by campaign, tracking issues found and settings changed. Then rank fixes by expected impact and risk.
A typical order of work after an audit:
- Fix tracking and duplicates first, because every later change depends on them.
- Correct conversion values and primary goals.
- Consolidate campaigns that cannot reach enough conversions.
- Add missing exclusions and refresh customer lists.
- Fill creative gaps.
- Only then adjust bidding targets, in small steps.
Signs an account needs an audit now
You do not have to wait for a scheduled review. Some symptoms suggest the automated systems are working from bad signals and need attention before more money goes in.
- Conversions in the platform rise while sales or qualified leads in the backend stay flat.
- Cost per conversion jumps after a site update, a new checkout or a tag manager change.
- Performance Max spend shifts heavily toward display or video placements with few sales.
- Campaigns stay in learning for weeks after every small edit.
- Different team members have changed targets repeatedly without a written log.
Each of these points back to the checklist above. Treat them as triggers for at least Steps 1 and 2.
Where to start this week
Take your largest ad account and complete Step 1 only. List every primary conversion action, compare 30 days of conversion counts with your backend or CRM and note the gap. If the gap is large, fixing it will likely improve automated performance more than any bid change. We run the same checklist at the start of every account engagement, and the tracking review alone usually takes less than a day.
Keep reading: AI Advertising Optimization · AI Advertising Optimization