The complete PPC audit checklist

Set the frame before you open a single campaign.

• Fix the window

  90 days is the standard read: long enough to show patterns, current enough to matter. Note anything anomalous inside it, a sale event, a stockout, an outage, before you judge the numbers.

• Write down the commercial target first

  An audit without a target is a tour. Derive the CPA or ROAS the business actually needs from margin, not from last year's dashboard, and judge everything against it.

• Confirm access and linked accounts

  Read-only access is enough to audit. Confirm Google Ads is linked to GA4 and Merchant Center, note where conversions are imported from, and map the account hierarchy before drawing conclusions.

• Pull a baseline

  Spend, revenue or leads, blended ROAS or CPA, impression share, and the top twenty queries and products by spend. Findings only count if you can show the before.

Checked automatically by the free audit

Everything else in the account learns from this data. Audit it first: bad signal makes every later finding unreliable.

• One primary action per goal

  Smart Bidding optimises toward whatever is marked primary. Two primaries for the same outcome means double counting; a purchase firing from both the Google tag and a GA4 import is the classic case.

• Retire dead and duplicate actions

  Actions with no conversions in 90 days, legacy tags still firing, and page-view style actions marked primary all blur the signal. Fewer, cleaner actions beat many stale ones.

• Enhanced conversions on, offline outcomes in

  Enhanced conversions recover signal lost to browser privacy. For lead generation, import qualified or closed outcomes from the CRM so bidding learns from deals, not form fills.

• Send values, not just counts

  Revenue, or better margin, for ecommerce; scored values for leads. If every conversion is worth 1.00, the system cannot tell a good customer from a cheap one. Google's floor for value-based bidding: two or more distinct values.

• Check consent and tag health

  Consent Mode gaps in Europe now directly shrink the conversions your bidding sees. Browser-only tracking on a large account is a risk flag; server-side tagging is the fix at scale.

• Re-verify attribution and windows

  Only data-driven and last click remain. Confirm conversion windows match the purchase cycle, and re-check imports after GA4's April 2026 attribution change: conversions can silently redistribute across campaigns.

Structure decides what the algorithms can learn and what your reports can tell you.

• Isolate brand from non-brand

  In Search and in Performance Max, using brand exclusions. Blended together, brand inflates every average and hides the true cost of acquiring a new customer.

• Segment by objective and margin, not history

  Campaigns should map to how the business makes money: objective, margin band, geography. Structures inherited from old team org charts fragment budget and blur accountability.

• Consolidate starved campaigns

  Campaigns limping below roughly thirty conversions a month learn slowly and pay for it in the auction. In 2026, fewer strong campaigns beat many weak ones.

• Find overlap and cannibalisation

  Check search terms served by both Search and Performance Max. Demand Gen audiences that include people already searching for you claim credit Search would have won anyway.

• Naming, tagging and experiment hygiene

  Auto-tagging on, one UTM convention (lowercase, underscores, one format), and no forgotten drafts or experiments still spending quietly in the background.

Checked automatically by the free audit

Where budget caps delivery on profitable campaigns, demand you already earned goes to a competitor.

• Find impression share lost to budget

  Campaigns beating their target while losing impression share to budget are the cheapest growth in the account. Fund them before you fund anything new.

• Kill zombie budgets

  Small budgets scattered across many campaigns keep each one too starved to learn. Consolidate them behind the campaigns with proof.

• Check pacing against seasonality

  Seasonality adjustments and data exclusions are the sanctioned levers for known events under Smart Bidding. Confirm past sale periods were excluded, and the next one is planned.

• Question spend concentration

  Know what share of spend sits in the top five campaigns, and whether that concentration is a decision or an accident of history.

Checked automatically by the free audit

The most common finding in mature accounts is not waste. It is throttling: targets so conservative that Smart Bidding turns profitable demand away.

• Match the strategy to the data volume

  Value-based bidding wants at least 15 conversions in 30 days and two or more distinct values, Google's published floor. Below that, target CPA or maximise conversions is the honest choice.

• Calibrate targets to margin, not history

  A ROAS target inherited from an old dashboard quietly turns profitable demand away. Strong ROAS with high impression share lost to rank is the tell that targets are throttling delivery.

• Change targets with discipline

  One move per conversion cycle, in steps rather than swings. Google's own guidance: frequent target changes keep the bidder re-learning instead of performing.

• Strip the legacy controls

  Most manual bid adjustments are ignored under Smart Bidding. The levers that still work: conversion data quality, value rules, seasonality adjustments and hard exclusions at minus 100 percent.

Query expansion is wider in 2026 than it has ever been. Negatives are the guardrail.

• Read the search terms report weekly on major spenders

  With broad match common and Smart Bidding Exploration on by default, the account is reaching queries it never used to. Triage what it found: scale the good, negate the bad.

• Use match types deliberately

  Broad match earns its place where tracking is clean and volume supports it. Exact and phrase keep control on the money terms. The mix should be a decision, not sediment.

• Maintain negative keyword discipline

  Shared negative lists, applied to Performance Max as well. Brand negatives inside non-brand campaigns. A competitor policy that is deliberate rather than accidental.

• Hunt duplicates and conflicts

  The same query served from multiple ad groups splits data and bids against itself. Negatives that block your own keywords are more common than anyone admits.

• Read Quality Score as a symptom, not a target

  Low ad relevance or landing page experience components point at real gaps worth fixing. Optimising the number itself is a distraction.

Checked automatically by the free audit

The platforms compose ads from whatever you feed them. The audit question is whether the inputs, and the controls, are deliberate.

• A strong RSA in every ad group

  Ad strength good or better, distinct headlines rather than fifteen rephrasings, and pinning only where compliance forces it. Pinning caps what the system can compose.

• Complete the asset stack

  Sitelinks, callouts, structured snippets, images, business logo. Eligibility for AI-composed surfaces depends on the assets you feed them; gaps are invisible lost auctions.

• Review AI Max before September does it for you

  AI Max is GA and Dynamic Search Ads auto-upgrade from September 2026. Set brand controls, locations of interest, URL exclusions and natural-language instructions deliberately, before the migration.

• Approve what the machine wrote

  Google's systems now generate and select creative by default, and AI-generated content labelling rules apply. Review the automatically created assets report; you remain responsible for every asset that serves.

• Watch creative freshness beyond Search

  Demand Gen and Performance Max now expect active creative management: formats for Shorts and in-feed, video coverage, refresh before fatigue shows in frequency and falling engagement.

Checked automatically by the free audit

For retailers, the catalogue is the campaign, and this is the heart of any ecommerce PPC audit. Most accounts monetise a fraction of it.

• Clear disapprovals and limited items

  Merchant Center first: policy issues, missing identifiers, price and availability mismatches between feed and site. Every disapproved product is demand you cannot serve.

• Measure catalogue coverage

  What share of eligible products received an impression in 90 days? Eligible products with zero impressions are a structure problem, not a demand problem.

• Know your revenue concentration

  If the top twenty products carry the account, that should be a strategy, not a surprise. The long tail is where visibility, and incremental growth, is usually lost.

• Treat the feed as an AI surface

  Titles, attributes and availability now feed AI answers as well as Shopping. Feed quality is simultaneously a paid and an answer-engine play; accuracy compounds.

• Structure Performance Max for control

  Asset groups by margin or category, brand excluded, placements reviewed, negatives applied, and the channel-level reporting Google now provides actually read each month.

Checked automatically by the free audit

Waste hides in the tail. So do the winners being starved to pay for it.

• Cut spend with nothing to show

  Products and queries burning meaningful click spend with zero conversions across the full 90 days. Rank by spend, decide deliberately: fix, demote or exclude.

• Feed the hidden gems

  The mirror image of waste: products returning five times or more on minimal spend are starved winners. Redirect spend from the noise to the proven drivers.

• Audit placements and partners

  Display and Demand Gen placements, search partners, app inventory. Exclude what spends without converting; keep the exclusion lists shared and current.

• Check geography and schedule

  Locations and hours with spend and no return. Confirm location settings target presence, not interest, where that distinction costs money.

The platforms differ; the discipline does not. Every channel bids on the signal you feed it.

• Rebuild the same signal spine everywhere

  Meta Conversions API with deduplication against the pixel; Microsoft UET with enhanced conversions. Match event coverage and values to what Google receives, or the comparison is fiction.

• Audit Meta structure like you audit PMax

  Advantage+ consolidates: fewer campaigns, stronger signals. With targeting automated, creative is the main lever, so check frequency, fatigue and refresh cadence.

• Stop blind-importing into Microsoft

  Imported campaigns inherit Google assumptions that do not hold. Review audience network placements, device performance and parity after every import cycle.

• Hold retail media to the same bar

  On Amazon and retail media networks: brand and generic isolated, share of voice tracked on hero SKUs, and incrementality asked of the channel like any other.

• Reconcile what the platforms claim

  Add each platform's reported conversions together and you will have more conversions than orders. Work from a blended view, and where spend justifies it, prove incrementality with a holdout test rather than trusting platform attribution.

In 2026, purchase decisions are shaped in AI Overviews, AI Mode, ChatGPT and Perplexity before a search query is ever typed. Demand now shows up in AI answers before it shows up in your search terms, so the audit has to look there too.

• Measure your AI answer share

  Run your money queries through AI Overviews, AI Mode, ChatGPT and Perplexity monthly. Log whether you appear, who is cited and who is recommended. You cannot manage a surface you have never looked at.

• Track LLM referral traffic

  Segment GA4 for traffic arriving from AI assistants (chatgpt.com, perplexity.ai, gemini.google.com) and watch its conversion rate. This traffic tends to arrive decided; make sure it lands well.

• Make your content citable

  Structured data (Product, Offer, FAQ, Organisation), specification and comparison pages that answer questions directly. Answer engines cite sources that commit to an answer.

• Confirm eligibility for ads in AI surfaces

  Ads in AI Overviews and AI Mode are served from the Search, Shopping, Performance Max and AI Max campaigns you already run. There is no separate buy; eligibility flows from asset completeness and feed quality.

• Read demand upstream of the keyword

  Prompts differ from queries: longer, more decided, later in the journey. Ask whether your planning would notice a demand shift in LLMs before it reaches search volume. The Rubin model reads demand signals across search, social and LLMs for exactly this reason.

The click is only half the transaction. The other half is everything after it.

• Test speed where the money lands

  Core Web Vitals on the top ten landing pages by spend, on mobile. The homepage passing means nothing if the product pages taking paid traffic fail.

• Check message match, and page selection

  Query to ad to page should read as one thought. AI Max now selects landing pages automatically, so exclude the pages that should never receive paid traffic.

• Walk the conversion path yourself

  Submit the form, place the test order. Confirm every tag fires once and only once, on the thank-you state, with the right value attached.

• Audit the consent experience

  A badly built consent banner suppresses the very conversions your bidding learns from. Balance compliance and signal; measure the consent rate, not just the banner's presence.