The practical concern is real: ChatGPT Search, Google’s AI search features, and Microsoft Copilot can surface web-derived answers and links before a buyer reaches a conventional search-results page. But the agency’s headline statistic, the claimed 31-to-74 improvement in 90 days, and the assertion that unmentioned firms are routinely losing business to named competitors remain proprietary benchmark claims, not independently verified market measurements.
Access Newswire published the announcement as a company release, and Morningstar’s version is a republication of that same wire material. Searches for independent reporting, a published methodology, or raw audit data did not turn up outside validation of the score, the client-loss claim, or the reported outcomes. For IT teams and marketers being asked to fund “AEO” work, that distinction changes this from an urgent market report into a vendor’s pitch built around a potentially useful diagnostic checklist.
The 31-Out-of-100 Figure Is Not a Neutral Industry Baseline
AI Search Engineers says every business in its audit sample had entity inconsistencies across platforms, while 94% lacked sufficient structured data, 91% had generic content not formatted for AI extraction, 89% lacked trusted-source citations, and 87% lacked documented outcome signals. Those figures sound precise, but the release does not identify the audited organizations, explain how they were selected, publish the scoring rubric in enough detail to reproduce results, or disclose the individual scores and variance behind each percentage.
The company’s own July 2026 material describes the AI Search Visibility Score as a 100-point framework divided into five 20-point categories: entity recognition, structured data, trusted-source citations, topical authority, and documented outcomes. That means the same agency determines what counts as a deficient signal, assigns the baseline score, implements its five-signal process, and measures the post-engagement result.
That does not make the framework fraudulent. It does mean the reported 31-point baseline and 43-point improvement cannot be read as an independently established industry condition or proof that the changes caused a corresponding increase in recommendations, leads, consultations, or revenue.
The sample is also too thin to support broad claims about professional services. More than 50 audits may be enough to reveal recurring implementation problems among prospects who sought an AEO agency, but it does not establish how law firms, medical offices, financial advisers, and consultancies at large perform in AI-mediated discovery. Businesses approaching an agency for visibility work are especially likely to have existing discoverability problems, creating an obvious selection bias.
The same issue applies to the nine completed engagements cited for the 31-to-74 score increase. The company explicitly says the data is internal, limited, unaudited, and not representative of all organizations. It does not say how many prompts were tested per customer, what geographic and service queries were used, which models and product configurations were involved, whether competitors changed during the test, or whether a higher score translated into a measurable increase in client acquisition.
Google Does Not Endorse a Separate AI-Markup Playbook
The biggest technical problem in the release is its treatment of structured data as though a stack of
FAQPage,
Review, service-specific,
LocalBusiness,
Person, and
ContactPointschema creates a direct route into AI-generated recommendations.
Google’s current Search Central guidance says the opposite on the central point: there are no additional technical requirements or special optimizations for appearing in AI Overviews or AI Mode. Pages need to be eligible for normal Google Search indexing and snippets, while Google recommends the same technical SEO fundamentals and helpful, reliable, people-first content it has long advised.
Structured data still has a legitimate use. It can help a search engine understand page content and, in supported cases, qualify content for particular search features. It should be deployed accurately where it describes material that actually exists on the page. But it is not an official admission ticket to Google AI Overviews, AI Mode, Gemini answers, ChatGPT Search, or Microsoft Copilot.
Some of the agency’s specific schema recommendations also need more caution than the release provides. Google sharply limited FAQ rich results in 2023, saying regular eligibility is now limited to well-known, authoritative government and health sites. A local law firm, financial adviser, or B2B consultancy can still use FAQ markup when it accurately represents a FAQ page, but it should not buy it expecting the historical FAQ expansion in Google results—or a guaranteed AI-search advantage.
Review markup has a similarly important limitation. Google’s documentation says self-serving reviews are ineligible for star-review treatment when the business being reviewed controls the reviews on its own
Organizationor
LocalBusinesspage. Adding a review widget or marking up customer praise on a firm’s own site may still be useful to visitors, provided it is truthful and compliant, but it is not a reliable path to a rich-result star display.
That does not invalidate entity cleanup, crawlability fixes, factual service pages, clear author information, or credible third-party references. Those are sensible web-publishing disciplines. The unsupported leap is from “these factors can improve a site’s clarity and discoverability” to “these are five verified signals that make AI platforms recommend a business.”
“ChatGPT, Gemini, and Copilot” Are Not One Search Channel
The release groups ChatGPT, Google Gemini, and Microsoft Copilot into one AI recommendation layer, then assigns a single score intended to measure business authority across them. That is a convenient sales model, but the products have different retrieval paths, controls, interfaces, and reasons for producing an answer.
OpenAI documents that ChatGPT Search can reformulate a prompt into targeted web queries and return an answer with sources when search is available and used. It does not promise that every recommendation prompt will search the web, return a recommendation, name a single provider, or cite the same sources across repeated runs.
Google draws a distinction between its Gemini product and AI features in Google Search. AI Overviews and AI Mode can display links to supporting web content, but Google says these experiences use different models and techniques, and the set of responses and links can vary. Google also says AI Overviews do not appear for every query, because they are shown only when its systems determine they add value beyond classic Search.
Microsoft’s documentation is even more explicit about the dependency on Bing and configuration. Microsoft 365 Copilot and Copilot Chat can ground answers in Bing web search, but administrators can manage or disable web search. Copilot may also rely on a user’s permitted Microsoft 365 documents rather than the public web, making a generic external “visibility” score a poor proxy for what a particular enterprise user sees.
A prompt test across these services is therefore a snapshot, not a durable ranking report. It can be useful for identifying errors—wrong addresses, stale professional biographies, nonexistent service descriptions, broken indexing, or missing authoritative pages. It cannot support a claim that one firm has a stable, portable “authority” score that predicts the output of multiple vendors’ changing AI systems.
The Client-Loss Claim Has No Conversion Evidence Behind It
AI Search Engineers’ sharpest assertion is that an AI system can name a competitor first and cause an unseen firm to lose the client before any other channel is encountered. In some high-intent queries, that is plausible: a buyer may accept a recommendation, click a cited source, or schedule a consultation without conducting wider research.
The release supplies no evidence that this is happening at a measurable rate among the audited firms. It does not report prompt-level recommendation frequency, referral traffic from AI platforms, branded-search changes, call tracking, consultation bookings, CRM attribution, conversion rates, or closed-won revenue. It also does not distinguish a business being absent because an AI answer gave no provider names from a business losing to a specific named competitor.
The traditional-search comparison is overstated as well. Google’s AI results frequently include multiple links and references, while ChatGPT Search and Copilot can cite sources rather than simply declare a single winner. In professional services, buyers also have licensing constraints, conflicts checks, insurance networks, geography, confidentiality requirements, fee structures, and referral relationships that an AI answer cannot settle.
For regulated professions, the safer operating model is to treat AI answers as one discovery surface among several, not as a replacement for reputation management, compliant advertising, local search, referral networks, and direct client due diligence.
The useful takeaway from AI Search Engineers’ release is narrower than its warning: firms should verify that public facts about their identity, locations, people, services, and credentials are consistent, crawlable, current, and supported by substantive first-party material. They should also test realistic buyer prompts periodically across the AI products their clients actually use, recording answers, cited sources, geography, logged-in state, and dates rather than relying on a single score.
But a 31-out-of-100 audit score is not evidence that a firm is invisible, and a 74 is not evidence that it has secured a place in AI recommendations. Until an AEO vendor can show reproducible prompt testing and connect visibility changes to independently auditable referral and conversion data, its score should be treated as an internal implementation metric—not a forecast of clients already lost.
References
- Primary source: ACCESS Newswire
Published: August 9, 2026 at 4:11 PM UTC
Loading…
www.accessnewswire.com - Related coverage: morningstar.com
- Related coverage: help.openai.com
Loading…
help.openai.com - Related coverage: help.openai.com
Loading…
help.openai.com - Related coverage: learn.microsoft.com
Data, privacy, and security for web search - Microsoft Copilot Studio | Microsoft Learn
Information on the Web Search functionality in Microsoft Copilot Studio regarding data privacy and security.learn.microsoft.com - Related coverage: blog.google
How AI Mode and AI Overviews help you explore the web
New updates to AI Mode and AI Overviews in Google Search make it easier for you to dive deeper online.blog.google - Related coverage: blog.google
Loading…
blog.google - Related coverage: developers.google.com
Loading…
developers.google.com - Related coverage: developers.google.com
Loading…
developers.google.com - Related coverage: learn.microsoft.com
Use public websites to improve generative answers - Microsoft Copilot Studio | Microsoft Learn
Learn how to structure generative answers nodes using public websites in Microsoft Copilot Studio. Optimize Bing Custom Search for accurate, grounded responses.learn.microsoft.com