Google AI Overviews: What's Actually Happening to Your PPC Budget

Google launched AI Overviews (initially as Search Generative Experience, then AI Overviews) as a way to summarize answers directly on the search results page. The intent was to reduce the number of clicks needed to answer a question. It worked. The downstream effects for paid search are real, and they do not look like the organic traffic collapse that publishers reported — but they are not harmless either.

This post covers what is actually happening to paid accounts, which signals to watch, and what adjustments matter.

What AI Overviews Changed in Search Results Layout

Before AI Overviews, a Google search result page for most queries had a predictable structure: ads at top, organic blue links below, sometimes a featured snippet or knowledge panel. Clicks distributed across the visible options.

AI Overviews changed that for a large share of informational queries. Now, for queries where Google determines an AI-generated summary can resolve the search intent, the AI panel appears at the top — sometimes above the first ad position — and synthesizes an answer with cited sources below it. Many users read the answer and do not click anything.

For text ads, the practical effect:

• On informational queries that previously drove top-of-funnel clicks, ad impressions may still show but CTR has dropped as the AI panel captures user attention before they reach the ad
• On transactional and commercial queries, Google continues to show ads with high frequency, and Shopping ads have appeared inside the AI panel for product searches
• Position 1 no longer guarantees the dominant visual prominence it once did on affected query types

For Performance Max and broad match campaigns, the effect is more pronounced because both campaign types target a wide query range — including informational queries now absorbed by AI Overviews.

The Three PPC Problems AI Overviews Create

Problem 1: Volume and Intent Shifts in Broad Match

Broad match captures queries the advertiser did not explicitly target. Before AI Overviews, this included a meaningful share of informational queries — "what is the best X," "how does X work" — that converted at low rates but generated top-of-funnel awareness.

AI Overviews now answer many of those queries in-place. The traffic that used to flow from those searches to ads and landing pages has shrunk. What remains is a higher-intent slice.

That sounds positive — less wasted spend on informational traffic. In practice, it creates problems:

• Volume drops that look like campaign degradation but are actually a structural market change
• CPA inflation because the same spend is reaching fewer users, even if those users convert better
• Smart Bidding disruption when the model was trained on a volume baseline that no longer exists

Accounts seeing unexplained impression share or volume drops in the past 12 months should check whether affected queries align with informational intent categories now consistently captured by AI Overviews.

Problem 2: Increased Automated Crawler Traffic

AI systems are trained on web content and continuously re-crawl the internet to update their knowledge bases. Commercial search interest in a product or service also drives AI research agents, price comparison bots, and content aggregators to crawl advertiser landing pages.

Some of these automated systems follow paid ad links. The behavior differs from human visitors:

• Zero or very short session duration
• No scrolling or interaction events
• Direct URL exit without navigating to other pages
• Repeated visits to the same landing page from different IPs within a short window
• No form engagement or conversion events

This traffic inflates paid session counts, raises bounce rate, and — most damagingly — registers in Google Analytics and Google Ads as low-quality paid sessions that Smart Bidding interprets as signals.

The scale of this problem is not uniform. Accounts in categories with high AI research interest — finance, healthcare, software, legal services, home services — see more of it. Accounts with high-value keywords attract more scraping because their landing pages are commercially valuable information.

Problem 3: Smart Bidding Signal Corruption

This is the least visible problem and the most damaging over time.

Google's automated bidding strategies — Target CPA, Target ROAS, Maximize Conversions — use historical conversion data to predict which future clicks are likely to convert. When the traffic mix changes, the model recalibrates. When the traffic mix includes a growing share of non-converting automated sessions, the model learns from those sessions too.

The corruption path:

This is not hypothetical. Any invalid traffic that reaches the landing page and registers in Analytics or via Google Ads conversion tracking has the potential to corrupt the signal Google uses to allocate budget and set bids.

The specific risk from AI crawler traffic is that it often shares characteristics with real users — similar geographic distribution, similar device types, similar time-of-day patterns — because AI crawling mirrors real human search interest. The Smart Bidding model cannot easily distinguish it from a weak human session.

Which Account Types Are Most Exposed

What to Check in Your Account

Check 1: Search Terms Report for Intent Shift

Pull 90 days of Search Terms data. Filter for queries you are paying for that contain "how to," "what is," "best," "vs," or "review." These are informational signals. Check whether:

• Impressions on these terms dropped over the past 6 months
• CTR on these terms dropped even when impressions held
• Conversion rate on these terms changed

A drop in impressions suggests the query pool shrank — AI Overviews may be resolving those queries in-place. A drop in CTR with stable impressions suggests ads are showing but not being seen or clicked because the AI panel is absorbing attention.

Check 2: Session Quality on Paid Traffic

In Google Analytics 4, segment paid search sessions and check:

• Bounce rate (or engagement rate) trend over the last 12 months
• Average session duration trend
• Sessions with 0 events versus sessions with meaningful engagement

A rising share of paid sessions with near-zero engagement that are not filtering through Google's invalid click detection is a crawler or non-human traffic signal. It may not show up in Google Ads invalid click columns — those detect clicks, not what happens after the click.

Check 3: Conversion Rate Variance by Query Type

If you have campaign segmentation between high-intent and broad-intent keywords, compare conversion rate trends. If broad-intent conversion rate has declined but high-intent is stable, the informational traffic that used to fund top-of-funnel is changing character — lower volume, lower intent, or more automated.

Check 4: Smart Bidding Performance Benchmarks

Pull Target CPA or Target ROAS performance over 12 months. If the model has been consistently over-spending to hit targets, or if recommended target adjustments have become more frequent, check whether the training window includes a period of high mixed-intent traffic. Bidding models trained on a wider query mix from 18 months ago may not reflect current query character.

Adjustments That Help

Tighten negative keyword lists for informational queries. Add negatives for query patterns consistently resolved by AI Overviews in your category. This reduces spend on impressions with declining CTR and removes the AI-crawler-magnet landing pages from the serving path.

Segment campaigns by intent more aggressively. Separate transactional keywords (buy, price, quote, near me) from informational keywords. This makes it easier to isolate where AI Overview impact is showing up and bid accordingly.

Add invalid traffic filtering before conversion events register. Crawler and bot sessions that reach paid landing pages should not register as conversion events in Google Ads. ClickFortify monitors for non-converting automated traffic patterns and filters sessions that match crawler behavior before they pollute Smart Bidding signal.

Review conversion import settings if using offline conversions. If your CRM is importing leads back to Google Ads as conversion events, ensure fake or spam leads from confused AI-driven traffic are filtered before import. Bad lead data in Smart Bidding is harder to correct than bad click data.

Watch impression share, not just clicks. If AI Overviews are absorbing queries, absolute impression share may drop even when your bid is competitive. Distinguishing a bid problem from a query-pool problem requires comparing Search Impression Share trend against Search Lost IS (Rank) and Search Lost IS (Budget) separately.

The Longer-Term Picture

Google has a commercial incentive to make AI Overviews work alongside ads, not instead of them. They have demonstrated Shopping ads inside AI Overviews, lead generation integrations, and business profile features embedded in AI panels. For transactional queries — the ones that drive ad revenue — Google is not reducing ad inventory.

The shift is on informational queries. PPC campaigns that depended on capturing top-of-funnel informational traffic cheaply through broad match have a structural problem that is not going to reverse. The query pool is smaller, the clicks that remain are higher-intent, and the automated traffic that fills the gap is not buyers.

The accounts that adapt fastest are the ones already running tighter campaigns: exact and phrase match on high-intent terms, aggressive negative keyword hygiene, and invalid traffic filtering that covers post-click behavior, not just pre-click detection. Those practices were good before AI Overviews. They are necessary now.