Google's New AI Reports and the Share-of-Voice Problem: How Agencies Are Starting to Measure AI Search Visibility

For most of the past two years, agencies trying to show clients how they appear in AI answers had to rely on third-party trackers and screenshots. That changed in a small but meaningful way this summer, when Google began giving site owners first-party data on their visibility in AI Overviews and AI Mode. The new reports answer one question well and leave several others open, and that gap is shaping how agencies are starting to report AI search to clients.

What Google Now Shows, and What It Doesn't

On June 3, 2026, Google announced new "Search Generative AI performance reports" in Search Console, with dedicated views for Search and Discover. According to the post by Hillel Maoz, a Search ecosystem engineering manager, and Moshe Samet, a product manager lead for Search Console, the reports show impressions, meaning how often a site's URLs appeared in generative AI features such as AI Overviews and AI Mode, along with which pages appeared, country, device (for Search) and date at hourly, daily, weekly or monthly granularity.

Google initially released the reports to a subset of sites. A note added to the post says that as of August 31, 2026, the insights have been rolled out to all websites worldwide. Google said the same data remains part of the overall performance report and that it may add "additional metrics over time."

What is missing matters as much as what is there. Search Engine Journal, reporting on the worldwide rollout on August 31, noted that the reports do not include click data, and that Google's help pages indicate some sites without sufficient AI impressions may not see them. Google also launched a control alongside the reports that lets site owners exclude their content from AI Overviews, AI Mode and Discover's AI features; according to SEJ, Google said the setting "is not used as a ranking signal elsewhere in Search." SEJ also reported that the UK Competition and Markets Authority has set a March 2027 deadline tied to page-level controls and click-through reporting.

For agencies, that means the most authoritative AI visibility number now available is an impression count for Google's own AI features, with no query or click detail. It says nothing about ChatGPT, Perplexity or Gemini.

Share of AI Voice and Its Measurement Problem

To cover the rest, agencies are turning to a growing set of tracking tools that run prompts through AI assistants and record which brands and sources appear. The metrics go by various names, including mention rate, citation share and "share of AI voice," but most boil down to how often a brand shows up across a set of prompts compared with competitors.

Research published January 28, 2026 by SparkToro co-founder Rand Fishkin and Gumshoe.ai co-founder Patrick O'Donnell shows why that has to be done carefully. In their study, 600 volunteers ran 12 prompts through ChatGPT, Claude and Google's AI a combined 2,961 times in November and December 2025. They found there is "a <1 in 100 chance" that ChatGPT or Google's AI will return the same list of brands in any two responses, and that "it's more like 1 in 1,000 runs before you'd see two lists in the same order."

The conclusion was not that AI visibility cannot be measured, but that ranking positions are the wrong unit. The authors found that "visibility % across dozens to hundreds of prompts run multiple times is a reasonable metric." They also urged marketers to ask AI-tracking vendors for transparent, public research on their methods.

Vendor benchmarks are appearing to give those percentages context. Semrush's 2026 AI Visibility Index, released June 26, 2026, analyzed 126 million US AI search prompts from January through April 2026 across ChatGPT, Gemini, Google AI Mode and Google AI Overviews, separating brand "mentions" in answers from "citations" of supporting domains. It found concentration varies sharply by industry: the top three brands held 82.9% of visibility in news and media, compared with 41.4% in finance. Semrush also reported that 45% of marketing leaders say they cannot accurately measure brand visibility in AI-generated answers. These are Semrush's own figures, and the company sells AI visibility tools.

Agencies Feel the Attribution Gap

Agency survey data shows the pressure. The 2026 Marketing Agency Benchmarks Report from AgencyAnalytics, which surveyed 494 agency professionals between February and April 2026 (64% based in the US), found that 64% named Google's AI Overviews a top industry concern and 66% reported increased client demand for SEO or optimization for AI search, the top new service category in the survey. On measurement, 48% said tracking users who discover brands through AI tools is their hardest attribution problem, and 42% said clients are asking why organic traffic is declining.

"Finding ways to report on how clients are showing up in AI results is going to be critical to our success," Rachel Li, marketing director at Swifty, was quoted as saying in AgencyAnalytics' July 2026 analysis of the findings. AgencyAnalytics sells client reporting software, which is worth keeping in mind, but the concern it describes is consistent with the Semrush finding.

What This Means for Your Business

For agencies, a defensible AI visibility report now has three layers. The first is Google's own Search Console generative AI impressions, reported by page and tracked over time, which provide a first-party baseline even without clicks. The second is a tracked prompt set covering the questions a client's buyers actually ask, run repeatedly across the assistants that matter for that audience and reported as a visibility or share-of-voice percentage against named competitors, not as a rank. The third is business outcomes: branded search, direct traffic, and leads that report finding the client through an AI tool.

Be explicit with clients about uncertainty. AI answers vary from run to run, tracking tools use different methods, and Google's report does not yet tie impressions to clicks. A trend line over months is more reliable than any single snapshot.

For business owners and in-house marketers, ask any agency or tool vendor how many times each prompt is run, which platforms are covered, and how prompts were chosen. A report that shows a brand ranking "number one in ChatGPT" without that context is, based on the SparkToro research, not telling you much.