A quick note before this week's issue: We've updated how we select, verify, and package The Visibility Report. We'd love your feedback on what is more useful, what's missing, and what we should improve.

The Visibility Report #32

Week of August 24 - August 30, 2026

We've been treating AI visibility as one score. This week's research shows what that misses: a brand can appear and earn citations while a competitor wins the recommendation.

Presence, readiness, and business impact need separate measures

Foundation tested Brex across 348 responses to six unbranded questions on eight AI engines. Brex appeared in 51.7% of answers, but brex.com supplied 6.5% of citations. Ramp appeared more often and won recommendations, even when Brex ranked higher in Google or supplied a cited page. In a separate 11-thread revenue-intelligence test, Gong appeared every time and won three.

Aleyda Solís's August 26 framework separates presence, readiness, and business impact. Presence covers how a brand appears, is cited, and gets recommended. Readiness diagnoses the structural causes; business impact tracks commercial value.

A blended visibility score can hide the competitor being chosen. Add a recommendation test to high-value prompts: ask the model to choose, compare, and explain why. Record the outside sources behind its verdict. Foundation measured one August window and notes that answers and source sets change, so use these results as a baseline rather than a permanent score.

Sources: Foundation's Brex analysis | Foundation's Gong test | Aleyda Solís's measurement framework​

Google is expanding some AI Overviews automatically

Google confirmed that some AI Overviews may open into fuller answers without a "Show more" click, pushing traditional results lower and adding an AI Mode-style follow-up box. Google says this happens when its systems judge the fuller answer useful. Expansion stops after someone starts scrolling so the page does not jump underneath them.

Barry Schwartz separately documented the fuller response and AI Mode prompt box loading without "Show more."

Track how the result renders alongside whether an AI Overview exists. The expanded version takes more space and changes the value of every citation inside it.

Sources: Search Engine Land | Search Engine Roundtable​

Microsoft opened AI Max to all Search advertisers

AI Max can broaden search-term matching, customize ad text, and select a different final URL. It uses keywords, ads, landing pages, and other campaign context to reach longer conversational searches, including the underlying searches an AI prompt may generate.

Microsoft says advertisers can inspect the new matches, assets, and landing-page routes. Existing budgets, bidding, conversion goals, targeting, and negative keywords still apply, with added brand and term controls.

Across 44 advertiser-run A/B tests, Microsoft reported a 13.6% spend-weighted conversion lift. Ten tests produced statistically significant positive results, so the aggregate is early vendor evidence, not a benchmark.

Run the suite against a clean control and review the query, creative, and destination together.

Source: Microsoft Advertising​

Claude and Claude Code retrieve different sources

Profound compared 24,135 responses across 1,724 prompts. With web search enabled, Claude searched in 93% of responses and Claude Code in 13%. Brand overlap for the same prompt averaged about 20%.

The products also visited different parts of websites. Nearly three-quarters of Claude Code's top page visits went to documentation, informational, and pricing pages. Claude spent more of its observed activity on robots.txt files, sitemaps, and home pages.

Profound sells AI visibility software, and its agent-traffic sample came from domains it tracks. The study did not report human review of its page classifications or test whether its recommended changes improved visibility.

Test Claude Code separately for technical buyers. Make documentation, compatibility details, and pricing easy to extract.

Sources: Profound's study | Search Engine Journal's review​

AI Mode is adding travel transactions

Google added flight price tracking, points and miles comparisons, and hotel booking through partners. Flight tracking is available in more than 180 countries and territories. Hotel booking is rolling out in U.S. English with Booking.com, Expedia, Hilton, Marriott, Priceline, and others.

Google is shortening the path from query to purchase. Visibility now depends on accurate inventory, structured product data, partner integrations, and a working handoff. Audit whether the engine can verify price, availability, policy, and the next action.

Source: Google​

Reddit leads Google's forum results

Ahrefs analyzed 145 million U.S. results. Google's Discussions and forums feature appeared in 11.7%. Reddit showed in 87.8% of those U.S. results and 83.9% worldwide. Facebook Groups followed at 38.3%, ahead of Quora at 32%.

The study covers Google's forum feature rather than AI citations. Those same community pages can still influence answer engines and buyers. Find the communities already visible for customer questions and contribute material worth retrieving.

Source: Ahrefs​

From the Tool Blogs

Profound: Its Claude versus Claude Code study shows why measurement should stay product-specific.

Profound: New Adobe Experience Manager nodes let Agents list, search, read, create, and update Content Fragments. The changes can be staged for review before publication.

Semrush: Promptwatch found Reddit's share of ChatGPT citations fell from 3.8% to 0.5%. The firm calls the result provisional, which is another reason to track source shifts rather than treat one channel as permanent.

Metehan Yeşilyurt: He separately found a 4.6%-to-0.23% drop alongside increased site:-restricted searches. Different methods and short windows make this directional corroboration, not confirmation.

Semrush: Its refreshed AI Overview tracking guide separates mentions, cited pages, triggering queries, and competitor visibility.

BrightEdge:Agent Edge serves AI agents a lighter page while human visitors receive the original.

From Practitioners

Search Influence:New AI SEO Tools to Watch in 2026 reviews KIME, LLMrefs, ZeroChannel.ai, and Promptwatch, including methodology, data sources, refresh rates, and recommendations.

What to do next

• Track appearance, owned citation share, and recommendation separately for prompts tied to revenue.
• Test products such as Claude and Claude Code separately, even when they share a model family.
• Map the comparison pages, reviews, publishers, and communities that shape answers outside your site.
• Check whether AI systems can verify price, availability, policy, and the next action.
• Review AI Max query expansion, generated creative, and destination changes together against a clean control.

Test five high-intent prompts where you already appear. Ask the model to choose an option and explain why. Record the winner, cited sources, and objections.

The Visibility Report | Will Scott​
This newsletter is produced collaboratively by Will Scott and Bob, an AI agent. Human oversight, AI efficiency.
​Unsubscribe​