Most small business owners don’t need another marketing dashboard. They need a straight answer to three questions: What changed, why did it change, and what should we do next?
Google is trying to answer those questions inside Google Ads and Google Analytics. On August 10, the company announced new AI summaries, prompt-built reports, personalized insight cards, and peer benchmarking. The tools could save hours of report building. They could also make a shaky recommendation look more certain than it is.
The right way to use them is as a fast analyst, not an automatic manager. Let the AI find patterns and assemble evidence. Keep budget, targeting, creative, and conversion decisions with a person who understands the business.
What Google Actually Added
The announcement covers several related features across the two platforms.
Google Analytics now places AI Overviews at the top of its homepage. These summaries call out changes since your last visit, such as a traffic decline or seasonal sales increase. A card can pass its context into Ask Advisor for follow-up analysis. Google also says users can receive these summaries by phone or email at a chosen frequency.
Google Ads has a revised homepage with AI-generated insight cards tailored to the account. A prompt box lets you request a custom insight, such as an explanation for lost impression share or a review of recent campaign trends.
New Google Ads Dashboards can turn a plain-language prompt into a visual report. Google says each report also generates a real-time explanation of what may be driving the result. Similar dashboard functionality is planned for Google Analytics.
Finally, Ask Advisor in Analytics can compare campaign performance with anonymized averages from similar businesses. Google’s existing GA4 benchmarking documentation explains that peer data is presented using the median, 25th percentile, and 75th percentile. That is more useful than a single vague “industry average,” but it still needs context.
Availability may differ by account. Google’s help page describes Ask Advisor as an in-product conversational tool built with Gemini and identifies it as a beta. If you don’t see every feature yet, check again rather than rebuilding your reporting process around a promised button.
Start With Measurement, Not Prompts
An AI summary can’t repair bad tracking. If a lead form fires twice, phone calls aren’t imported, or internal employees appear as prospects, the system will explain a distorted picture with impressive confidence.
Before using the new reports for decisions, verify the handful of events tied to money. For a service business, that usually means qualified form submissions, booked appointments, calls over a meaningful duration, and completed purchases or deposits. A click on an email address may be useful diagnostic data, but it isn’t equal to a signed contract.
Also confirm that Google Ads is optimizing for the right actions. Google’s optimization score documentation says advertisers can choose which conversions are included for optimization and can override an inferred strategy with objectives such as target CPA or target ROAS. If every button click is marked as a primary conversion, both the bidding system and the AI analyst are working toward the wrong goal.
Use this quick preflight check:
- Submit every important form and confirm one conversion appears, not zero or two.
- Call each tracked number and verify that the call source and duration arrive correctly.
- Compare purchases, booked jobs, or qualified leads in your CRM with Analytics totals for the same period.
- Label primary conversions as revenue-related outcomes and keep softer engagement events secondary.
- Record major promotions, price changes, site releases, outages, and sales staffing changes on a simple marketing calendar.
That last item matters. Google can see that conversion rate dropped on Tuesday. It may not know that your estimator was out sick, your online booking calendar had no openings, or a storm shut down the shop.
Five Prompts Worth Using First
Open-ended requests such as “How are my ads doing?” tend to produce broad answers. A useful prompt names the period, comparison, metric, segment, and desired output.
Try these as starting points:
- “Compare qualified lead volume and cost per qualified lead for the last 28 days with the previous 28 days. Separate brand and non-brand campaigns. Show the three largest changes and the evidence behind each explanation.”
- “Find campaigns where spend increased by more than 15% while qualified conversions stayed flat or declined. Show the dates when the change began.”
- “Build a weekly dashboard with spend, qualified leads, cost per qualified lead, conversion rate, and lead value. Break results out by campaign and device.”
- “Compare mobile and desktop landing-page performance. Flag pages with at least 100 sessions where mobile conversion rate is materially lower than desktop.”
- “Compare our campaign performance with the relevant peer group. Show the median and percentile range, then identify which differences are actionable and which may reflect business model or geography.”
The thresholds in those examples are operating choices, not universal standards. A business receiving eight leads per month shouldn’t demand a 100-session minimum for every segment. A store processing thousands of orders should use tighter filters. Set the threshold high enough to avoid reacting to two random clicks and low enough to catch a real problem before it gets expensive.
Ask for evidence every time. A helpful answer should identify the campaigns, date range, metric values, and comparison behind its explanation. “Competition increased” is a hypothesis. A documented loss in impression share, rising click cost, and unchanged conversion rate make it a testable hypothesis.
Read AI Explanations as Leads, Not Verdicts
AI reporting is good at compressing a large account into a short list of anomalies. It is not automatically good at causation.
Suppose an overview says paid search revenue fell 18% after mobile traffic shifted toward a new landing page. That correlation deserves attention. It does not prove the page caused the decline. The same week might include a holiday, a budget cap, an inventory shortage, or an offline sales problem.
Use a three-step review:
First, verify the numbers in the underlying report. GA4 can hide or approximate portions of a report under some conditions. Google’s data quality guidance tells users to inspect the data quality indicator for thresholding, sampling, or other limits. If a report is thresholded, don’t present its exact small-segment numbers as settled fact.
Second, look for a competing explanation. Check promotions, seasonality, inventory, staffing, tracking releases, and site changes. Search the Google Ads change log as well. Google says Change history retains account, campaign, and ad group changes for two years, so you can match a performance shift to a bid, budget, targeting, or asset change.
Third, choose a reversible test. Move a small amount of budget, repair one landing page, exclude a clearly irrelevant query group, or test new copy in a controlled campaign. Don’t reorganize the whole account because one summary card sounds persuasive.
Benchmark Without Chasing the Average
Peer benchmarking can answer a useful question: Is this result unusual for businesses that resemble mine? It cannot answer the more important question: Is this campaign profitable for my business?
Two plumbing companies in the same market may have different margins, close rates, service mixes, and capacity. One can profitably pay $180 for a qualified water-heater lead. The other may lose money above $70. A peer median can’t settle that difference.
Google explains that GA4 benchmarks use peer groups based on industry category and other business details, with aggregated data protected by minimum property counts and contribution limits. For some absolute metrics, Google estimates a range using the property’s active-user count. That makes the comparison directional, not an audited profit benchmark.
Use peer data to find questions:
- If acquisition cost sits near the 75th percentile, is weak conversion tracking, expensive traffic, or a poor landing page responsible?
- If conversion rate is high but volume is low, are campaigns too restricted or is the market simply small?
- If engagement looks strong but sales are weak, is the site attracting researchers instead of buyers?
Your internal numbers still win. Gross profit per job, lead-to-sale rate, refund rate, customer capacity, and cash flow determine what you can afford. Treat peer performance as a flashlight, not a target.
Put Approval Rules Around AI Recommendations
The phrase “agentic capabilities” matters because these tools are moving closer to action. That makes a written approval policy useful even for a two-person marketing team.
Let the tool summarize performance, create draft reports, group anomalies, and suggest tests. Require human approval for budget increases, bidding changes, location expansion, new broad-match targeting, conversion-goal changes, and automatically created assets that customers will see.
Keep automatic recommendations under review. Google states that when advertisers enable automatically applied recommendations, selected recommendation types can be applied regularly. Check which types are enabled, who approved them, and whether they match the business goal. Convenience is not a reason to give a system an unlimited operating range.
A simple rule works: any change that can materially raise spend, broaden the audience, alter the offer, or redefine success needs a named owner and a rollback plan.
A 30-Minute Weekly Workflow
These tools are most valuable when they shorten a consistent review rather than trigger constant tinkering.
Spend the first five minutes reading the AI Overview and listing only material changes. Use the next ten minutes to open the underlying reports, check data quality, and compare the dates with your marketing calendar and Change history. Take another ten minutes to ask focused follow-up questions and select one or two tests. Use the final five minutes to document the owner, expected result, review date, and rollback condition.
Judge the workflow on business output. Did it catch wasted spend earlier? Did it reduce time spent assembling slides? Did it help the team run a better test? If the weekly meeting becomes a tour of colorful charts, simplify it.
Google’s new AI reporting can make a small team faster. It cannot supply clean tracking, profit margins, customer context, or judgment. Get those inputs right, ask narrow questions, verify the evidence, and keep approval over consequential changes.
If your Ads and Analytics accounts produce more confusion than decisions, start a conversation with Your Web Team. We’ll help connect the reporting to the website, leads, and revenue that actually matter.
