Visio: from 10 to 193 appearances across four AI engines in 2026

Appearances in June 2026

10135 questions measured across 4 engines, 540 measurements in the month. Appearance is the broad measure: recommended, mentioned in the text, or with the URL cited as a source; it includes questions that already name the brand.

→

Appearances in July 2026

193the same 135 questions, the same 4 engines, judge and rubric. July’s raw data is not in murmur’s repository: this number is read from the indexing report.

period

The comparable window is June to July 2026, using the same 135 questions, judge and rubric at both endpoints. Before it is a 23 May 2026 snapshot, valid as a qualitative baseline of invisibility, not as a line-by-line comparable number.

company size

It does not appear in an approved source. Visio’s brand profile gives category, headquarters and founding year, but no company size.

universe

135 questions comparable between June and July, worded identically in both cycles and measured across 4 engines, for 540 measurements a month. In June, each question ran 5 times, totalling 2,700 responses: 2,560 open-search and 140 prompted, which name the brand. The measured wave was Portuguese: 135 questions out of 834 planned. The English (505 questions) and Spanish (463) waves remain pending, and the paragraph that gives the three slices does not state how they add together, so they are not added here.

engines

Four: ChatGPT, Claude, Google (AI Overview) and Perplexity. The May snapshot is treated as three engines, and even then yielded data from two: Google returned empty in collection and Perplexity was not yet measured.

AI named a competitor in 95% of responses and Visio in none of them, in June 2026

Visio sells operations software for retail networks, franchises and food service: margin, cost of goods sold, store-level reconciliation and loss prevention. Its domain is visio.ai, and its Brazilian operation is in São Carlos, in the interior of São Paulo. What is measured here was measured by murmur.marketing, a Brazilian Generative Engine Optimization operation run by Mateus Gomes and hired by Visio for this work: it is the same organisation that executes and measures, which is why every number on this page has a standard, denominator and date.

In June 2026, murmur measured 2,700 AI responses about Visio: 135 buyer questions across four engines, five repetitions each. Of the 2,560 responses where the question did not supply the brand name, Visio was recommended zero times. Zero across all four engines, and zero citations of its site as a source. In the same set of 2,560, 95% of responses named a competitor.

That zero and the scorecard’s 10 measure different things, and it is worth stating which is which before continuing. The 10 counts appearance in 540 measurements—135 questions run in each of four engines—and includes questions that already name the brand. The zero counts spontaneous recommendation in the 2,560 responses where the question did not provide the name. June’s few appearances were almost all on the prompted side: in Perplexity, the only 6 were all of the ‘Visio vs X’ type, with the name in the question.

Before that was an older snapshot, from 23 May 2026, valid as a qualitative baseline rather than a line-by-line comparable number: collection used an earlier format with fewer questions, Google returned empty, and Perplexity did not run. In questions that did not name it, the brand appeared in 0 of 37 Claude responses and about 2 of 37 ChatGPT responses. Those two were income-statement and accounting questions, probably about the namesake.

The cycle diagnosis was specific and guided everything that followed. The problem was not a lack of published content. It was that AI did not recognise that name as a company in that market.

The wall had a name: Microsoft Visio

Of June’s 2,700 responses, 140 came from questions that provided the name Visio, 35 per engine. They were discarded from the main standard because supplying the name asks about the brand instead of testing whether AI brings it up independently. But those 140 count something else.

In 58 of them (41%), AI answered about Microsoft’s diagramming software, not the retail company. In 32 (23%), it recommended the retail Visio: 17 in Perplexity, 12 in ChatGPT, 3 in Claude (the report records Claude’s reading as inconsistent) and 0 in Google. Google did not recognise the retail company in any of the 35 questions it ran with the name included.

That changed the work and the standard. Publishing more would not solve it alone; AI had to connect that name with retail operations. Measurement therefore required an extra step, written in Visio’s brand profile before this page existed: a response counts as the brand only when there is a retail anchor in the question or response, or the visio.ai domain in citations. Without that, the judge marks a namesake and does not count it.

From 10 to 193 appearances between June and July 2026

The rigorous comparison is June against July 2026, across the same 135 questions, worded identically, with the same judge and rubric in four engines. That is 540 measurements per month. Visio appeared in 10 of them in June, and 193 in July.

No June appearance was lost: the regression column is zero in all four engines. Perplexity rose from 6 to 92, Google from 0 to 61, ChatGPT from 3 to 30 and Claude from 1 to 10. Perplexity alone accounts for 86 of the 183 gains, almost half. Google’s starting point was absolute zero, meaning none of the 135 questions brought up the brand in June.

The question types that opened up are those that matter to a seller: alternatives to X, competitors of X and generic categories—in other words, questions where the buyer does not yet know the brand.

A provenance caveat follows this figure in every piece, and it is a release condition, not decoration: July’s raw data is not in murmur’s repository. The indexing report, not the capture file, supports 193. June’s cycle has retained, auditable raw captures; July’s does not.

The Indexing API lost to the control group: 4 of 16 versus 7 of 16

murmur requested indexing for 16 URLs through Google’s Indexing API and left another 16 in the sitemap only, as a control group, to measure whether the push did anything rather than assuming it did. On 12 July 2026, the API-pushed group had 4 of 16 indexed. The control group, which received no push, had 7 of 16. The next day’s re-audit returned the same verdict. The report opens by saying indexing started—11 of the 32 audited pages—but it does not support the claim that the accelerator accelerated.

Most of July’s gain is source citation, not endorsement. In Claude, all 9 monthly gains are of that kind: Visio’s URL is cited, but the text does not select it as an option. In Google, about 42 of the 61 gains are sources and only about 19 are recommendations written in prose. The queue of what remains closed is large: of 135 questions, Visio remains absent in 74 in Google, 105 in ChatGPT and 125 in Claude.

The namesake wall also survived published content. In July, on the question ‘checklist facil vs visio’, Perplexity answered that Visio is better for drawing flows, diagrams and processes. None of this is in a footnote: it is in the same piece that records the rise, in Perplexity’s engine-by-engine reading, under the label ‘Risk to watch’; the absences return in a bullet called ‘Honesty about limits’, written by the person who performed the measurement.

English, Spanish, July’s raw data and any business outcome were left out

The measured wave was Portuguese: 135 questions out of 834 planned. The English (505 questions) and Spanish (463) waves did not run, and the report does not state how these three slices add together, so they are not added here. Visio is global, so this is a Portuguese snapshot, partial to the whole, and none of its conclusions applies to other markets.

The number’s two endpoints do not have the same degree of evidence, and that needs to be stated rather than discovered. June’s cycle has retained, auditable raw captures. July’s does not: July’s raw data is not in murmur’s repository, and the indexing report—not the capture file—supports 193. Anyone wanting to reopen both sides to the same depth cannot do so today.

There is an instrument difference the reports do not resolve, and it weighs most where the gain is largest. Both declare the same judge and rubric, but do not describe session mode in the same words: June records Perplexity running in a logged-in account, while July describes capture as anonymous. Because Perplexity accounts for 86 of 183 gains, this is no footnote detail. No one measured whether session mode changed between endpoints, and nothing in the repository supports saying it was identical.

Nothing in this case is a commercial outcome. murmur did not measure sales, leads, revenue or visits to Visio’s site. It also did not measure position within the response: there is no first-place figure for Visio anywhere in this work. Every figure above is a count of AI responses, and none means any of those other things.