Innovation Dreams Strategic AI / Case Study
Rogue Visual Design is a woman owned reality capture firm in Kansas City. In 2024 and early 2025, asking an AI system for LiDAR scanning services in Kansas City did not return them. Today, every one of the eight major AI products returns Rogue Visual Design first or near first for that question.
That is the headline, and on its own it is not very useful. The measurement underneath it is the part worth reading.
Between September 10 and September 11, 2026, we ran roughly 180 queries across eight AI products plus Google, in four separate rounds, using clean accounts and controlled conditions. The results were not uniformly good. They showed that visibility is not a property a business has or lacks. It attaches to specific words. Change the phrasing of the same request, describing the same service, in the same city, to the same buyer, and the result moves from every system finding the company to no system finding it.
There is no settled name for this discipline yet, which is itself a sign of how new it is. The same practice is currently published under at least six labels.
Generative engine optimization (GEO) is the most established, and now has its own encyclopedia entry. Answer engine optimization (AEO) is used nearly as widely, sometimes as a synonym and sometimes to mean the narrower work of being quoted directly in an answer. AI optimization (AIO), sometimes written as artificial intelligence optimization, tends to describe the off-site reputation and trust signals that make a system willing to recommend a business rather than merely list it. Large language model optimization (LLMO) circulates in agency writing and generally means the same thing as GEO. AI search optimization and AI visibility are the plain language versions, and they are what most business owners actually type.
Innovation Dreams Strategic AI uses AI visibility, because it describes the outcome rather than the technique, and because a business owner asking whether artificial intelligence can find their company is asking a clear question that does not require an acronym.
The terms are listed here for a practical reason. A reader looking for this kind of work may search any one of them, and a document that only uses one label is retrievable for only one label. That is the same finding this case study documents, applied to itself.
Innovation Dreams Strategic AI exists because of a question that came up while working on Rogue Visual Design's marketing. It was an ordinary one, and an old one. Why is a firm doing good work not getting found by the people looking for it?
That question has had the same handful of answers for twenty years. Wrong keywords. Weak backlinks. Slow site. Bad listings. This time none of them fit, and the answer turned out to be something that had not existed the last time anyone asked it.
I watched a business I cared about become invisible almost overnight. When I asked why they were not bringing in new clients I realized the real problem was AI. They had become invisible to it. Vanessa Murphy, founder, Innovation Dreams Strategic AI
The discovery came during an ordinary late night, crawling the company's own website to find out why it would not rank. What came back was not poor performance. It was nothing. No pages. No projects. Not even the company name.
The strange part was that all of it was right there. Opening the site in a browser showed every page, every project, the company name on every screen. Asking an AI system for 3D LiDAR scanning services in the area returned other firms. Asking about Rogue Visual Design by name produced an answer, but the answer was stitched together from fragments found elsewhere on the web rather than from the company's own site.
Pasting a direct link to a specific page did work. The system would read it and answer accurately. But it only held for that page, in that moment. Every follow-up question meant pasting the link again. There were ways around it, and they took enough iterations that no actual buyer would ever get there. Someone comparing vendors asks a question, asks two or three follow-ups, and moves on. Nobody troubleshoots a company's visibility on the company's behalf.
That is the moment the answer stopped being hypothetical. It was never about what a person could see on that website. It was about what artificial intelligence could not, and the gap between those two things was invisible from the browser.
The work that followed started narrow, with schema markup and the technical basics. It did not stay there. Every layer that was checked turned up something else that machines could not read, and the scope widened until the entire public footprint had been audited and rebuilt around one question: can a machine find this, retrieve it, and describe it correctly without help.
That question is what this practice is built on. If artificial intelligence cannot see a business, how can any AI powered system recommend it?
Innovation Dreams Strategic AI launched in June 2025. Work on the Rogue Visual Design website began that September, and the twelve months described below run from there to September 2026.
The Internet Archive holds independent snapshots of the previous Rogue Visual Design website. The snapshot dated July 20, 2025 is the closest complete capture before the site was rebuilt, and it is the reference point for everything that follows. It is third party hosted and was not produced by us.
The site was not neglected. It was well designed, it loaded, it described the work accurately, and it carried real project photography. That is exactly why it is a useful example. Invisibility to AI is not the same thing as a bad website.
Laser Scanning. 2D Floor Plans. Point Clouds. UAV/Drone. Projects. Blog. About. Contact.
Rogue Visual Design is a Kansas City based Reality Capture firm specializing in 3D Laser Scanning (LiDAR), Aerial Drones (UAV), and CAD/BIM modeling services. rogue3d.com homepage, archived July 20, 2025
Read that alongside the results in part five and the pattern becomes visible before we ever get to the numbers. The words in the navigation and the homepage description are the words AI returns the company for today. The words that are absent are the ones it does not.
The AI visibility work for this client ran continuously across the engagement. Website content and structure, publishing, contact routing, infrastructure, and third party presence were all in active work, most of them concurrently and most of them revisited more than once.
What follows is the work completed, grouped by what it changed rather than by date. Several items took repeated passes. The website was rebuilt, then audited again after structural defects survived the first launch. Publishing ran, stalled on an approval and image bottleneck, and was rebuilt. Crawler access was verified, then found blocked again by a different mechanism. Those reversals are part of the record and they are why the list below is organized by outcome instead of by calendar.
Several streams were active across the whole engagement rather than at any single point, and they are the reason the work above produced results instead of sitting inert.
Other work was delivered for this client across the same period that is not visibility work and is not counted here.
Three separate protective defaults on the hosting platform, across four months, each hid something from machines. None of them were errors by the client. They were defaults, switched on by the platform, doing what they were designed to do. For most websites that is harmless. For a firm competing to be found by AI it is the opposite of what is wanted, and none of it is visible to a human looking at the site.
Any site Innovation Dreams Strategic AI deploys now gets a documented pass through crawler rules, managed robots settings, and the full protection section before it is called live.
Every result in this document was produced under conditions designed to remove the most obvious ways a test like this can fool itself.
The automated tests and the consumer products disagreed, repeatedly and in one direction. One provider's automated interface missed a query that the same company's consumer product answered correctly twice in the same hour. Another provider's automated access was unavailable entirely on billing grounds while its consumer product answered cleanly with accurate detail.
Automated testing is a conservative floor, not the real picture. Consumer products turn search on by default in ways the developer interfaces often do not. Any visibility measurement built only on automated queries will understate real world results rather than overstate them.
Asked for laser scanning or LiDAR services in Kansas City, using the terminology the industry uses, all eight products returned Rogue Visual Design in a leading position, with accurate supporting detail including certification status, drone certification, federal registration, and the mix of local and national work.
Two of the eight went further than placement and made an explicit recommendation, telling the reader where to start. The other six listed the company accurately and prominently without endorsing it. That distinction is maintained throughout this document.
Eleven services, each asked as an open question about the best providers in Kansas City, with no company named.
| Service | Found | Reading |
|---|---|---|
| Laser scanning | 3 of 3 | Core identity, fully established |
| LiDAR | 3 of 3 | Core identity, fully established |
| Scan to BIM | 2 of 3 | Strong |
| Point cloud registration | 2 of 3 | Strong |
| As built documentation | 2 of 3 | Strong |
| Scan to CAD | 1 of 3 | Real gap, service is performed and published |
| UAV drone mapping | 1 of 3 | Real gap, two products listed ten or more local drone operators without this one |
| Orthomosaic mapping | 1 of 3 | Real gap |
| Photo documentation | 1 of 3 | Partial |
| Topographic mapping | 0 of 3 | Not a gap, see below |
| Aerial photography | 0 of 3 | Not a gap, see below |
Aerial photography was run deliberately as a control, because it was unclear whether this belongs on the service list at all. All three products built a category populated by cinema drone crews, real estate listing photographers, and marketing video shops. Rogue Visual Design does not belong in that category, and the miss is evidence that the instrument is measuring something real rather than reporting absence everywhere.
Topographic mapping returned a consistent market definition across all three products: a licensed surveyor and a civil engineering firm, with one stating plainly that a drone deliverable is not a legally sealed survey. That is a scope question for the business, not a content problem.
Four phrasings. The same service. The same city. The same kind of buyer. The only variable is word choice.
| How the buyer asks | Found |
|---|---|
| Scan to BIM provider Kansas City | 7 of 7 |
| Subcontractor for site documentation scanning Kansas City | 4 of 6 valid |
| Point cloud to Revit model Kansas City | 2 of 7 |
| Existing conditions survey firm Kansas City | 0 of 7 |
An architect, an engineer, and a general contractor asking for the identical service, in three different but entirely standard ways, get three different outcomes. One phrasing produces a clean sweep. Another produces nothing at all.
On the shutout, every product understood the request correctly. Each one returned a list of local firms. None of them included this one. Two named specific competitors as their recommendation. This is not a case of the systems misreading the question.
The phrase "existing conditions survey" does not appear on the archived site. The homepage came within one word of it, describing the work as documenting "the as-built, real-world conditions needed by design professionals and decision makers," and used a different word.
The phrase was never published, so it was never learned, so it is not returned. The gap is not mysterious and it is not expensive to close.
We also ran four phrasings a property owner would use, someone who has never heard the words LiDAR or scan to BIM. Those results ranged from none of eight to four of eight. Two distinct causes stacked there. In some cases the products misread the intent entirely, reading a request about scanning a building as a request about consumer 3D printing, or reading a question about historic buildings as a question about world heritage preservation projects. In the cases where intent was read correctly and the company still did not appear, that is a genuine content gap of the same kind described above.
One product was the consistent outlier across every plain language round, returning the company for three of four, including one unprompted statement that it would be the first call for a serious architectural or renovation project. The others ranged from none to two.
When asked about the company by name, one AI product stated that Rogue Visual Design has been in business since 2011, and cited the company's own About page as its source.
The archived About page does not say that. It says this:
Rouge Visual Design's first scanning project was in 2011. rogue3d.com About page, archived July 20, 2025
A first project date, read as a founding date. The sentence is accurate and the inference drawn from it is not. Meanwhile the homepage of the same site, in the same capture, described the firm as having been in business for over seventeen years, which points to a different year entirely.
Two dates, published simultaneously on one website, neither of them stated as a founding year. That is the origin of the competing dates that circulated across AI systems and business data providers, and one product surfaced the contradiction directly to a prospective buyer inside an otherwise positive recommendation.
The correct year is 2007, confirmed by the owners. On August 14, 2026, nine instances across five files were corrected and deployed, covering structured data, meta descriptions, body copy, and the About page. One remaining discrepancy sits on a personal profile owned by the client rather than on any surface under this engagement's control, and it was identified as theirs to correct rather than quietly changed.
What makes this worth dwelling on is the cause. Not neglect, not bad information, and not stale cached data. One sentence that a human reads correctly and a machine does not, published beside a second sentence pointing at a different year. Correcting the source is straightforward. What takes longer is the correction propagating into systems that already learned the wrong version, which is why the September testing still found the old date in circulation weeks after the fix shipped.
The archived About page opens with the company name misspelled, as Rouge rather than Rogue, in the first sentence of the section describing what the company is. Entity resolution in these systems runs on name matching. A misspelled company name on the page that carries the company's history is a mechanical reason for that content to fail to connect to the company it describes.
All of the above anchors on Kansas City. The firm also states, and the archived homepage stated in July 2025, that it serves clients across the United States. We tested that claim directly with a purpose built protocol: eleven questions in three groups, with no city, state, or nearby phrasing anywhere, each run twice, once with location unset and once with location pinned to Kansas City, so that the difference between the two is a measurement rather than an assumption.
| Question type, no location given | Found |
|---|---|
| Credential led, for example woman owned and federally registered providers | 6 of 16 |
| Project type led, for example military installations, preservation, stadiums, hospitals | 2 of 16 |
| Buyer situation phrasing | 0 of 12 |
The reading is direct. The company is indexed as a Kansas City firm and not much else. Certifications are the only claims travelling outside the city. The national project history is not doing retrieval work. Buyer situation phrasing is a complete shutout, which reproduces the vocabulary pattern from part six at national scale.
Seven questions changed outcome between the two location conditions, five of them toward a hit once Kansas City was switched on. Two moved the other way, so the effect is directional rather than clean.
This is the clearest single lesson in the document. The national claim was published on the website for years. Publishing a claim and being retrievable for it are different things, and only one of them can be measured.
Every product asked would recommend Rogue Visual Design. Two without reservation, one positively, one conditionally. None of the reservations concerned the quality of the work.
The conditional one gave its reason, and it is the single most actionable sentence produced by the entire exercise. Everything it could find was company self description, directory listings, and employee profiles. No independent client reviews. No third party evaluation.
That is an authority gap stated in one line. A business can control its own website completely and still be described by AI as unverified, because everything the machine found traced back to the business itself.
The footer on every page published the company inbox address wrapped in an automatic protection feature that replaces the address with a script based decoder link. A human sees the address. A crawler sees nothing.
The consequence was demonstrated, not theorised. One AI product, while recommending the firm first for a subcontractor query, supplied the correct phone number and then offered a personal email address it had found on social media as the contact route, because the company address was invisible to it. Prospective clients arriving through AI were being pointed at a personal inbox.
The phone number returned by AI systems was independently verified as correct against the site and third party listings. The email address was not correct, for the reason above. Contact details returned by an AI system should be verified before they are trusted, in either direction.
Possibly, and possibly only for some ways of asking. In this engagement the same company, same service, same city scored a clean sweep across every AI product for one phrasing and returned nothing at all for another. The only way to know is to test multiple products with multiple phrasings, in clean sessions, and record the results.
The three most common reasons found here were all mechanical. A hosting platform setting was blocking AI crawlers by name without anyone switching it on deliberately. The words buyers actually use were never published on the website. And the only information available about the company came from the company itself, with no independent third party source, which causes some systems to describe a business as unverified.
Appearing in a list and being recommended are different outcomes and should be measured separately. In this engagement, placement was broad while unprompted recommendation language appeared on a minority of products for cold queries. When systems were asked about the company by name, recommendation was much stronger, with the one reservation given being the absence of independent reviews rather than anything about the work. Third party signal appears to be the difference between being listed and being endorsed.
A structured test of which questions cause AI systems to return a business and which do not, run across multiple products in controlled sessions, with the results traced back to specific causes on the website, the hosting platform, and third party sources. It produces a list of fixes, not a grade.
It overlaps and it is not the same. Traditional search optimization competes for a ranked position on a results page. AI visibility determines whether a system retrieves and describes a business at all when generating an answer, often with no click involved. A site can be technically sound for conventional search and still be invisible to AI, which is what was found at the start of this engagement.
AI answers vary between runs and change as models and indexes update. Results here are dated to a specific two day window for that reason, with a recheck planned at four to six weeks using identical wording and controls. A single spot check on a single product is not a measurement.
The multi provider tracking tools used in this engagement were built by Innovation Dreams Strategic AI and are reusable for any client. Rogue Visual Design was the proof case.