Last month in Sacramento, a woman walked into a three-bedroom ranch house she was about to buy for $610,000. Her home inspector, hired through a Yelp search, spent two hours and forty minutes walking the property, snapping photos, and tapping on things. He charged her $425, and she received a 47-page report the next morning, complete with color-coded severity ratings and a section on the HVAC system that described the condenser as "functional" despite a visibly corroded refrigerant line. She negotiated $2,000 off. Felt good about it.
Nobody asked the inspector for a license, and nobody could have, because California does not require one.
That list reads like a typo. California, the state that requires 1,000 hours of supervised training before someone can legally cut hair, demands exactly nothing from the person tasked with evaluating the structural integrity of a half-million-dollar home. Colorado, which licenses massage therapists at 500 hours and estheticians at 600, lets anyone with a tape measure and a flashlight call themselves a home inspector. Georgia requires a license to arrange flowers commercially but not to assess whether the foundation of a house is cracking.
This is not an oversight but a decades-old regulatory gap, and a generation of AI-powered inspection tools is about to collide with it in ways that could either fix the problem or make it significantly worse.
What the Numbers Say About What Inspectors Miss
A Clever Real Estate survey cited by the National Association of Realtors found that 65 percent of new home buyers who hired an inspector discovered issues, and 24 percent of new homes failed their first inspection outright. But that statistic only captures homes that were inspected at all. One in seven new home buyers skipped the inspection entirely. That is 714,000 homes a year. For existing homes, where defects accumulate over decades, the numbers are worse.
Repair Pricer analyzed 50,000 inspection reports and found more than one million repair items averaging over $11,000 per home. Most commonly: 55 percent of homes had doors that needed adjusting, which is often an early indicator of foundation settlement. Fifty-four percent lacked exterior caulking, exposing them to water infiltration. Forty-eight percent lacked GFCI protection near water sources, an electrocution risk that has been required by the National Electrical Code since 1971 for bathroom outlets and expanded steadily since then.
Those are the problems inspectors found and documented. Nobody can answer the question of how many they did not, because state-level data on inspection miss rates does not exist in any systematic form.
Where Licensing Fails and Where It Doesn't Exist
In the 36 states that do require a license, requirements vary so widely that the word "licensed" communicates almost nothing about competence. Texas mandates 194 classroom hours plus 40 supervised field inspections before a license is granted, New York requires 140 hours of coursework along with passage of the National Home Inspector Examination and a full year of supervised practice, and Florida demands 120 hours of training and supervised inspections before anyone may legally charge a homebuyer for an opinion about their property's condition. South Dakota technically requires licensure but sets requirements so minimal that the barrier barely registers. The range is absurd.
Closer to a standardizing force is the National Home Inspector Examination, the NHIE, a multiple-choice test accepted by roughly 30 states. But passing a multiple-choice test does not mean someone can recognize the difference between a cosmetic stain on a basement wall and active efflorescence indicating persistent hydrostatic pressure. That distinction is experience, not curriculum, and it is precisely the kind of tacit knowledge that no multiple-choice exam can assess, because it accumulates only through hundreds of crawl-space hours spent staring at joists and feeling the give of subfloor under your boots.
Two professional associations, InterNACHI and ASHI, maintain their own certification programs. InterNACHI dwarfs the field with 1.6 million website visits in the first four months of 2025 alone, compared to 35,900 for ASHI, according to SEMRUSH data published by InterNACHI. InterNACHI's Certified Professional Inspector designation requires passing an exam and completing continuing education. ASHI's standards are arguably more rigorous but reach far fewer inspectors, which means that the dominant credentialing body in American home inspection is a voluntary association whose certification requirements, however well-intentioned, carry no legal weight in the states where quality control matters most. Neither body has the power to revoke anyone's right to practice in an unlicensed state, which means the entire quality-control apparatus for home inspection in California, Colorado, Georgia, and eleven other states rests on Yelp reviews and word of mouth.
Where AI Enters the Crawl Space
Into this regulatory vacuum, a wave of AI-powered inspection tools is arriving, and the timing is not coincidental. A home inspection software market worth $3.2 billion in 2025 is projected to reach $5.6 billion by 2033. Verified Market Research predicts that 74 percent of structural inspections will eventually be assisted by computer vision. These tools are no longer speculative; they are shipping to paying customers, and the companies building them are raising money on the premise that every inspection workflow can be partially or fully automated.
Binsr Inspect, a Phoenix startup that raised $1.1 million in seed funding from New Stack Ventures, is the most explicit about what it is trying to accomplish. Its platform lets inspectors narrate findings by voice while walking a property: "Crack in the HVAC condensate line, garage, maintenance item." AI transcribes the note, categorizes it, selects the appropriate comment template from a library of thousands, grabs a still frame from the video feed, annotates the image, and drops it into the correct section of the report. A process that currently takes inspectors two to four hours of post-walkthrough desk work collapses into something that finishes on-site.
Mark Garcia, Binsr's co-founder, told AZ Big Media that the average inspector generates 150 to 300 line items per report. Over a career, that accumulates to roughly 8,000 stored comments, buried in folder hierarchies that inspectors manually search during every walkthrough. "It's a grueling process," Garcia said. Binsr has about 50 active customers and more than 500 companies on its waitlist, all of this with a team of three people.
Other tools push the boundary between augmentation and outright replacement. Spectin lets users upload inspection videos and photos, then generates a PDF property condition report with severity rankings and code relevance flags, no inspector required. Home Inspector Buddy, which launched on Product Hunt this year, analyzes photos to flag potential issues and rank them by severity. These consumer-facing tools position themselves as pre-screening aids, something a buyer might use to evaluate a property before committing to a professional inspection. In practice, they are a short step from replacing the professional entirely, and nobody in the regulatory apparatus is paying attention to the difference because nobody in the regulatory apparatus regulates the profession in those fourteen states.
What AI Could Fix, and What It Could Fake
Consider two scenarios for the same technology. In the first, a veteran inspector in a licensed state uses Binsr or a similar tool to cut report-writing time in half, freeing her to inspect a third home each day while maintaining the same thoroughness during walkthroughs. AI handles the transcription, categorization, and formatting while she continues to identify the problems herself, and productivity rises without sacrificing thoroughness because the technology eliminated the desk work, not the judgment calls that actually protect buyers. Price pressure may even drop, making inspections more accessible to the buyers who need them most.
In the second scenario, someone in California or Idaho with no formal training downloads Spectin, walks through a property filming with their phone, uploads the footage, and receives a professional-looking report that flags peeling paint but cannot identify the reason the second-floor bathroom subfloor feels soft to the touch. Color-coded severity ratings and code references fill the buyer's screen, and the report looks every bit as credible as one produced by a twenty-year veteran who actually knows where the load-bearing walls are. Nobody in the transaction knows that the person who produced it could not tell a load-bearing wall from a partition.
Both scenarios are already happening, neither has a technological safeguard, and the same tools serve a twenty-year veteran and a first-day amateur with identical interfaces, identical outputs, and identical professional formatting.
This is a version of a problem that policy scholars call "credentialism arbitrage." In licensed states, AI tools augment a professional whose baseline competence has been verified by the state. In unlicensed states, the same tools provide the appearance of professionalism without the underlying knowledge. What separates a Binsr-augmented veteran from a Spectin-equipped amateur is enormous, a gap measured in thousands of hours of field experience, hundreds of crawl spaces, and a vocabulary of defects that cannot be downloaded but must be accumulated one leaking flashing, one cracked foundation, one improperly wired subpanel at a time. Yet their reports may be indistinguishable to a buyer scrolling through a PDF on their phone during a lunch break.
What Missed Inspections Actually Cost
Home inspections are not academic exercises. A typical inspection in the United States costs between $300 and $500 for a service that takes two to four hours. Roughly five million existing homes change hands annually, and the majority involve an inspection, which makes this a roughly $2 billion industry.
Missed defects dwarf the cost of inspections themselves. Not slightly. By orders of magnitude. Repair Pricer's $11,000 average repair figure, applied across even a conservative estimate of the miss rate, suggests billions in post-closing repair costs that might have been negotiated differently or avoided entirely with a more thorough evaluation. 2-10 Home Buyers Warranty reports that the average structural warranty claim runs $70,000 and typically surfaces four to seven years after closing. Foundation repairs alone range from $2,200 to $8,100 nationally, with severe cases exceeding $20,000.
In an unlicensed state, a buyer has limited legal recourse if the inspection was negligent, because licensed states typically require inspectors to carry errors and omissions insurance at a minimum of $150,000 in many jurisdictions, while unlicensed states require nothing, which means a buyer in California, the most expensive housing market in America, has less statutory protection from a bad inspection than a buyer in Texas, where the median home costs a third as much. A buyer's only protection is a clause in the inspection contract, which most inspectors cap at the fee they charged. Four hundred twenty-five dollars of liability protecting a $610,000 purchase. That is the deal.
What Would Actually Help
The useful policy intervention is not to ban AI inspection tools, which would be both impractical and counterproductive, given that the tools genuinely improve efficiency for qualified inspectors. What would actually help operates on two tracks.
First, the 14 unlicensed states should adopt minimum licensing standards. An existing national exam, the NHIE, is already accepted by the majority of licensed states. Requiring passage of that exam plus a minimum number of supervised field inspections, as Texas and New York already do, would establish a floor without creating an onerous barrier to entry. California's insistence on licensing barbers at 1,000 hours while requiring nothing of home inspectors is not merely absurd, it is not merely a regulatory curiosity that makes for a good headline, and it is not merely an embarrassment for a state that prides itself on consumer protection legislation. It is a quantifiable transfer of risk from the state to individual home buyers who have no way to evaluate the competence of the person they are hiring.
Second, AI tools should be required to disclose their role in the inspection process. If a report was generated partially or entirely by AI analysis rather than human observation, the buyer should know. Not because AI analysis is inherently inferior, it may in fact be more consistent for certain categories of visual defect, but because the buyer is entitled to understand what they are paying for. A human inspector who missed the corroded refrigerant line made a judgment error born of fatigue, distraction, or insufficient training, while an AI that was never shown what a corroded refrigerant line looks like has a training gap born of whatever images its developers chose to include in the dataset and whatever images they did not. Different failures demand different responses. And buyers deserve to know which one they got.
Limitations of This Analysis
State-level data on inspection miss rates does not exist in any systematic way, which makes it impossible to compare outcomes in licensed versus unlicensed states with statistical rigor. McKissock Learning's 14-state unlicensed count reflects regulatory status, not necessarily the absence of voluntary credentialing. InterNACHI members in unlicensed states may be more competent than licensed inspectors in states with weak requirements. Assumed, not proven, this correlation between licensing and quality is the central weakness of the regulatory argument.
Every AI tool discussed here is early-stage, and none has been independently studied for its impact on inspection quality, accuracy, or the downstream financial outcomes that ultimately matter to the buyers whose largest purchase depends on the results. Binsr's waitlist and customer numbers are self-reported. Verified Market Research's 74 percent computer vision prediction is a forecast, not a measurement, and market research firm predictions in technology adoption are historically unreliable.
Why Licensing Might Not Be the Answer Either
Licensing requirements can function as guild protections rather than consumer safeguards, and the Institute for Justice, a libertarian legal organization, has long argued exactly this point. Their research shows that states with more stringent licensing requirements do not consistently produce better-quality services across a range of professions, which, applied to home inspection, suggests that California's lack of a licensing requirement may not be the primary driver of poor inspections. Competence comes from experience, mentorship, and market reputation, not from passing a test. InterNACHI's voluntary certification program, which has far greater reach than any state licensing board, may be a more effective quality mechanism than a government license.
That counterargument has real merit, particularly given that the NHIE tests knowledge recall rather than field judgment. But it breaks down at the margin: in the absence of any licensing requirement, the buyer has no minimum quality signal at all. Even a low bar is a bar, and even a floor can be built on.
Catherine Chen covers policy, regulation, and the law as it applies to residential construction. She has written about AI-automated permit review, zoning algorithms, and the insurance implications of robot-built homes.