# Where Is the Real Opportunity in Pet Tech?

> Pet spending hit $158 billion, but the strongest pet-tech opportunities may lie in what general AI cannot create on its own: physical automation, trusted human supply, proprietary measurements, and real-world care workflows. My most-used pet app is ChatGPT, which is the whole problem for pet chatbots.

**Published:** 2026-08-27
**Canonical URL:** https://yihuisong.com/article/pet-tech

My cat, Jojo, just turned two.

For her birthday, I started thinking about what else I could get her. Technology has already quietly entered her life. At home, she has an automatic feeder, a circulating warm-water fountain, a dry-food rehydration feeder, an automatic litter box, and an AirTag on her collar in case an indoor cat somehow escapes.

For someone who builds and thinks about AI products constantly, this list is surprisingly boring. None of these products tries to understand what my cat is thinking. None needs a clever AI interface before it becomes useful. They feed her, give her water, clean up after her, or help me find her.

That made me wonder why a market this large and emotionally intense has produced so few pet-tech products that feel like obvious startup opportunities.

A friend who works in U.S. pet consumer products told me the category is still concentrated in traditional segments like food, treats, and toys, where innovation is often incremental. Another friend, formerly in VC, gave me a much harsher framing: the pet industry is brutally difficult, many public companies look weak, and the only defensible areas tend to involve healthcare, regulation, biotech, or infrastructure.

I decided to see which version was closer to reality.

## A Huge Market Harder Than It Looks
The U.S. pet industry is still growing. Pet spending reached $158 billion in 2025, up 3.7% from the prior year. 95 million American households owned pets, including 71 million with dogs and 53 million with cats. Cat ownership alone grew 5%, with the [American Pet Products Association](https://americanpetproducts.org/news/u.s.-pet-industry-reaches-158-billion-in-2025-poised-for-continued-growth-in-2026) attributing part of that increase to Gen Z and Millennial adoption.

The growth looks more modest once pricing is taken into account. APPA estimates that roughly two percentage points of projected 2026 growth will come from higher prices rather than higher volume. It also reported that 22% of pet owners spent less on their pets in 2025, suggesting growing price sensitivity even as the overall market continues to expand.

The unevenness becomes clearer when looking at public companies.

[Chewy](https://investor.chewy.com/news-and-events/news/news-details/2026/Chewy-Announces-First-Quarter-2026-Financial-Results/default.aspx) reported Q1 FY2026 sales up 7.7%, with active customers growing 3.6%. The more revealing figure was composition: Autoship, Chewy's subscription-like recurring purchase program, represented 84.4% of net sales, and Autoship customer sales grew 10.5%. Scale combined with deeply habitual purchasing still works.

[Petco](https://ir.petco.com/news-releases/news-release-details/petco-reports-first-quarter-2026-results/) told a different story. Q1 FY2026 sales grew 0.2%, with comparable sales up 0.7%. Management described services as a "key engine" of growth. The contrast suggests that services were contributing more momentum than traditional merchandise.

[BARK](https://s27.q4cdn.com/974260903/files/doc_financials/2027/q1/Q1-FY7-Earnings-Release.pdf)'s Q1 FY2027 revenue fell 23.4%, with direct-to-consumer revenue down 25.2% and total orders dropping from 2.8 million to 2 million. Management cut marketing spend as part of a shift toward profitability. Falling revenue and order volume suggested that strong emotional attachment to pets does not automatically translate into durable subscription demand.

One structural difference between Chewy and BARK stood out to me. Chewy's Autoship turns repeated purchases of food and other consumables into a habit. BARK's toy boxes had to earn every renewal against the question of whether the dog actually needed another one.

That distinction extends well beyond these three companies. Many of the easiest products to launch, including treats, toys, collars, and accessories, are also the easiest to substitute. They compete primarily through formulation, design, branding, retail placement, and customer acquisition cost. As a cat owner, I see this constantly. Similar toys, feeders, accessories appear across dozens of brands with only superficial differences, and switching costs are usually close to zero.

Recent startup funding reinforces that distinction. The larger rounds are concentrated in veterinary care, clinical software, biotech, insurance, compliance, and other businesses built around deeper infrastructure or regulated workflows. Connected hardware continues to attract capital, but the disclosed rounds are generally smaller. Generic pet chat, journaling, and social products are largely absent from the biggest financings in my 2023–2026 sample.

![A directional 2023 to 2026 funding census by archetype. The largest disclosed rounds cluster in veterinary care, biotech, insurance, and clinical software, while connected hardware trails and pet chat, journals, and social products are largely absent.](/images/pet-tech/visual1.jpg)

What interested me was which technology had actually earned a permanent place in owners' routines, the products that became useful enough to stop feeling like technology at all.

## The Pet Tech That Worked Starts With a Job the Owner Already Has
### Start With the Job, Then Earn the Data
When I looked at the pet products I kept using, a simple pattern emerged. Each one solved a job I already needed done.

My own stack makes the point quickly. The automatic feeder keeps meals on schedule. The fountain keeps water circulating. The rehydration feeder handles a preparation step I used to do manually. The litter box scoops itself. The AirTag is there in case something goes wrong. Each product solves an existing chore or risk. None of them asked me to care about data before they became useful.

[Whisker](https://www.whisker.com/litter-robot-5-pro)'s Litter-Robot shows how this pattern can deepen. Its original job is straightforward: automate litter scooping. The current product also tracks weight, visit frequency, waste classification, and individual cats over time. Whisker's [health-data positioning](https://www.whisker.com/clean-living) emphasizes changes in weight or visit patterns that an owner may want to raise with a veterinarian. The sequence matters: a behavior the cat already performs, automated chore, passive data, then longitudinal health signals. Collecting that data requires no new habit from the cat or the owner. The cat was already going to use the litter box.

### Some Jobs Still Depend on Trusted Human Supply
[Rover](https://www.blackstone.com/news/press/rover-agrees-to-be-acquired-by-blackstone-in-2-3-billion-transaction/?utm_source=chatgpt.com) shows what happens when the underlying job cannot be automated. When an owner travels or works late, someone still has to walk, feed, or supervise the animal. Rover makes that labor easier to find, vet, book, pay for, and hire again. By September 2023, Rover had [facilitated more than 93 million services](https://www.sec.gov/Archives/edgar/data/1826018/000182601823000053/ex991_20231106.htm?utm_source=chatgpt.com) for more than 4 million pet parents. In its last public quarter, repeat bookings represented 84% of total bookings, and adjusted EBITDA margin reached 26%. [Blackstone acquired the company](https://www.blackstone.com/news/press/blackstone-completes-acquisition-of-rover/?utm_source=chatgpt.com) for approximately $2.3 billion in 2024. Rover's demand comes with hard deadlines, offline consequences, and transactions, which helps explain the repeat behavior.

[Modern Animal](https://investor.chewy.com/news-and-events/news/news-details/2026/Chewy-to-Acquire-Modern-Animal-Accelerating-Evolution-to-a-Fully-Integrated-Healthcare-Ecosystem/default.aspx) shows how the same logic can extend into veterinary care. It combined 29 physical clinics, membership, 24/7 virtual care, and proprietary software, serving more than 100,000 member families when Chewy announced the acquisition in April 2026. Chewy expected Modern Animal to add [more than $125 million in annualized run-rate revenue](https://investor.chewy.com/news-and-events/news/news-details/2026/Chewy-to-Acquire-Modern-Animal-Accelerating-Evolution-to-a-Fully-Integrated-Healthcare-Ecosystem/default.aspx), and agreed to acquire the company for approximately $400 million. The acquisition is a strong validation of a model where software supports a job that still requires people and physical infrastructure.

For my indoor cat, automation covers a larger share of the daily routine than it can for a dog. Feeding, water, litter, and cameras can reduce how much human intervention is needed while I am away. But cats still create exceptions that automation cannot handle, especially when a longer trip happens or a cat needs special care. Rover's [2026 acquisition of Meowtel](https://www.rover.com/blog/press-release/meowtel/?utm_source=chatgpt.com), a cat-sitting marketplace with more than 4,000 active sitters that had served more than 125,000 cats, suggests that trusted human intervention still matters even when much of the routine care can be automated.

### Safety Can Become a Sensing Layer
The pet-tracking category shows a different version of the same progression. [Tractive](https://tractive.com/blog/en/press/tractive-launches-new-cat-and-dog-trackers-and-features)'s original job was simple: find a lost pet. Its current product measures activity, sleep, territory, [resting heart rate](https://help.tractive.com/hc/en-us/articles/25432430590226-What-is-Resting-Heart-Rate-monitoring), and respiratory rate, learning individual baselines over time. For my indoor cat, an AirTag already handles the escape-risk concern with less friction, so a dedicated tracker has to justify its size, charging, and subscription through additional health value.

The history of Whistle also shows the pressure toward scale. Mars [acquired Whistle for more than $100 million](https://techcrunch.com/2016/03/31/whistle-acquired-by-mars/) in 2016 and later [connected its behavioral data with Banfield veterinary records](https://www.mars.com/news-and-stories/press-releases/mars-pet-insight-project). In 2025, [Tractive acquired Whistle's customer base and technology assets](https://tractive.com/blog/en/press/tractive-acquires-whistle), and the standalone [Whistle platform shut down](https://www.whistle.com/blogs/news/whistle-joins-the-tractive-family). Connected hardware comes with devices, firmware,  apps, connectivity, subscriptions, support, and data infrastructure. That infrastructure makes scale matter in a way a simple software product does not.

The newest AI wearables push the idea further. [Traini](https://traini.app/) raised [$7.5 million in 2026](https://www.globenewswire.com/news-release/2026/2/20/3241730/0/en/traini-s-sentra-smart-collar-ranked-1-ai-hardware-product-at-ces-2026-company-secures-7-5m-to-expand-pet-emotion-ai-platform.html?utm_source=chatgpt.com) around a smart collar that combines vocalization, movement, physiological signals, and environmental context. I find the sensing direction more convincing than its marketing around dog-to-human "translation." Research suggests that machine learning can extract meaningful patterns from dog vocalizations, including context and individual identity. A [2024 peer-reviewed study](https://pubmed.ncbi.nlm.nih.gov/38648990/?utm_source=chatgpt.com) adds to that evidence. Literal dog-to-human "translation" requires a much stronger claim about semantic ground truth, which current research has not established.

![Whisker, Rover/Meowtel, Modern Animal, and Tractive start from different jobs, but repeated use turns each into proprietary context: an existing job, then repeated use, then proprietary context, then deeper value.](/images/pet-tech/visual2.jpg)

Across these cases, pet tech creates durable value when it attaches to a job the owner already has, either by removing the work or by organizing human labor that cannot be removed. Once a product earns repeated use, it can also gain something more valuable than the original convenience: proprietary data about the animal.

The next question is what happens when that data stops being merely convenient and starts becoming medically relevant.

## Health Is Where Uncertainty Becomes Expensive
Health changes the value of pet data because the consequences of uncertainty become much higher.

### When Uncertainty Becomes a Financial Problem
My own cat is still young and has had no major illness. I got her from a certified CFA breeder, so I knew quite a lot about her early life before she came home. My ownership so far has been unusually calm.

Some of my friends have had very different experiences. One friend's older cat developed cancer and went through chemotherapy. The pet insurance helped, but he still paid several thousand dollars out of pocket. Another friend's Maine Coon appeared healthy at its last annual exam, then developed an acute blood cancer. Repeated emergency visits, procedures, and hospitalization generated bills large enough that the owner had to borrow money. After that cat died, the same friend purchased two Norwegian Forest Cat kittens through a broker. One developed acute FIP shortly after arriving and could not be saved.

These individual stories cannot tell us anything about disease rates. They do expose a consistent problem: owners rarely know when an ordinary health concern is about to become an expensive emergency.

The financial data points in the same direction. In [Synchrony's 2025 Pet Lifetime of Care Study](https://www.carecredit.com/pressrelease/synchrony-2025-pet-lifetime-of-care-study/?utm_source=chatgpt.com), nearly eight in ten owners underestimated lifetime pet-care costs, and only 31% said they could comfortably manage a major pet expense. In a separate [specialty-care study](https://www.carecredit.com/providers/insights/insights-veterinary-specialty-care-study/?utm_source=chatgpt.com), 72% of owners were equally concerned about their pet's outcome and the cost of treatment. These surveys are sponsored by Synchrony, which owns the CareCredit veterinary financing product. Still, the financial pressure they describe is consistent with the experiences above.

Insurance can reduce some of that risk. [NAPHIA](https://naphia.org/industry-data/?utm_source=chatgpt.com) reported 7.6 million insured pets in North America by the end of 2025. But insurance coverage does not remove the uncertainty around the care decision itself. Owners can still be unsure whether treatment is necessary, what it will cost, how deductibles and coinsurance apply, when reimbursement will arrive, or how much they can afford at the moment. [Spot's partnership with Tractive](https://tractive.com/blog/en/press/tractive-partners-with-spot-pet-insurance) hints at one possible connection between insurance and continuous monitoring, although the integration is still early and there is no public evidence that wearable data affects underwriting or pricing.

My breeder experience is relevant in a different way. [CFA's buyer guidance](https://cfa.org/cat-talk/so-you-want-to-get-a-new-pedigreed-cat/?utm_source=chatgpt.com) emphasizes access to family health history, breed-specific screening, [genetic testing](https://cfa.org/cat-talk/genetic-testing-for-pedigreed-cats/?utm_source=chatgpt.com) where available, and an ongoing relationship with someone who knows the animal's lineage. My own breeder followed this rule by providing a pedigree, vaccination history, early-life photos, socialization context, and ongoing advice. That gave me more information about my cat's health before I became her owner. However, it could not remove biological uncertainty entirely.

### The Businesses Built Around Reducing Uncertainty
For builders, reducing health uncertainty can create assets that are genuinely hard to reproduce. [IDEXX](https://ir.idexx.com/news-events/press-releases/detail/414/idexx-laboratories-announces-second-quarter-results) is the clearest example. Its Companion Animal Group diagnostic recurring revenue grew 10% organically in Q2 2026, VetLab consumables grew 14%, the inVue Dx installed base surpassed 9,000 instruments, and veterinary software and services revenue grew 12%.

IDEXX earns recurring revenue from consumables and diagnostics tied to installed instruments, while software embeds it deeper into veterinary workflow. The company is also roughly 40 years old with a massive installed base. It proves that this kind of moat can exist. A startup may not easily recreate a similar business.

A much younger company is testing a more consumer-facing version of the same principle. [Wonderdog](https://wonderdoghealth.com/) launched in 2026 with a $5 million pre-seed round and a preventive-health membership built around at-home blood collection. A veterinary technician draws the sample, Antech processes more than 40 biomarkers, a licensed veterinarian reviews the results, and the platform combines them with prior records to build a longitudinal baseline. An AI companion called Spark sits on top of that record.

Wonderdog creates the blood data Spark gets to reason over. A general-purpose AI would never receive that information unless someone first measured it. Creating the data requires Wonderdog to coordinate technicians, sample logistics, laboratory relationships, veterinary review, and repeat testing. The company is designed around earlier detection, but I did not find peer-reviewed validation showing that the platform itself improves clinical outcomes. It also remains unclear whether enough owners will repeatedly pay for preventive blood testing while their dogs appear healthy.

[PacaGen](https://pacagen.com/EN-US/pages/about-us) represents another route to defensibility. Its consumer products are built around a [protein-binder mechanism](https://pacagen.com/EN-US/pages/our-science) targeting household allergens, with supporting research that currently includes a [company-affiliated preprint](https://www.biorxiv.org/content/10.1101/2025.08.03.668213v1). I did not find independent peer-reviewed clinical validation of the finished consumer product. The proprietary science may make the product harder to copy, but it also raises the standard of evidence the company has to meet.

Pet health is where uncertainty becomes expensive. That is why it can support stronger willingness to pay and stronger product moats. The hardest assets to reproduce include measurements, diagnostics, validated mechanisms, clinical workflows, and insurance infrastructure. Those same assets increase validation burden, integration difficulty, capital requirements, and the consequence of error.

Yet a large amount of useful pet context sits outside those systems. Owners create it every day simply by living with the animal, and increasingly, that context can accumulate inside a general AI product.

## My Most-Used Pet App Isn't a Pet App
### AI Has an Unusually Low Social Cost
My most-used "pet software" is probably ChatGPT.

I chat about unusual behavior, photos I want interpreted, small health questions, stories about routines and affection, and plenty of moments that require no practical action at all. The relationship history accumulates there naturally through the photos, questions, and small observations I already bring to ChatGPT.

That raises a question I did not expect when I started this research: what can a pet-tech company know, measure, or do that ChatGPT cannot?

Several current pet apps are trying to own exactly the kind of context I have been putting into ChatGPT. [PetReveal AI](https://apps.apple.com/us/app/petreveal-ai/id6758315245), [PawLife](https://apps.apple.com/us/app/pawlife-pet-care-tracker/id6760197595), and [Fetch](https://apps.apple.com/us/app/fetch-pet-journal/id6757367776) combine photo-based analysis, symptom and behavior Q\&A, health timelines, journals, reminders, and personalized context. These are useful products. Much of that conversational surface now overlaps with capabilities available in a general multimodal assistant that can [understand images](https://help.openai.com/en/articles/8400551-image-inputs-for-chatgpt-faq.webp) and use [saved memories or past-chat context](https://help.openai.com/en/articles/8590148).

I share far more mundane cat photos and observations with ChatGPT than I would ever post publicly because there is almost no social cost to over-sharing with an AI. This behavior feels adjacent to [sharenting](https://pmc.ncbi.nlm.nih.gov/articles/PMC12344405/), the broader practice of parents repeatedly documenting and sharing their children's lives online. Pet owners do something similar with animals they treat as family: photographing small moments, narrating routines, and looking for reactions from others. [Research on pet-specific social media](https://journals.sagepub.com/doi/10.1177/1461444820956345) has documented the desire to share pets repeatedly, including the "fur baby" framing, the social politics of pet-photo posting, and the impulse to provide joy. With ChatGPT, I do not need another person to be interested enough to respond. I also no longer worry about whether anyone is tired of seeing my cat.

Pet-specific social apps approach part of this behavior by creating spaces where pet sharing and pet-centered interaction are expected. [Pawmates](https://apps.apple.com/us/app/pawmates-the-pet-social-media/id1397983772) offers local pet matching and playdates. [Petzbe](https://apps.apple.com/us/app/petzbe-pet-social-media/id1314000163) reports [more than 800,000 cumulative users](https://www.petzbeagency.com/portfolio/petzbe-app) on its pet-only social feed. These are real behaviors, but a standalone pet social app still needs enough local density, repeated participation, and a reason for people to use it instead of dog parks, Facebook Groups, Nextdoor, or Instagram. Rover has a structural advantage here because its demand comes with hard deadlines, offline consequences, and transactions. Pet friendship apps have optional intent, many substitutes, and harder local density requirements. Although pets give people plenty to talk about and connect over, that does not automatically make pet ownership a strong standalone social graph.

That pushes vertical pet products toward the things horizontal AI cannot generate on its own.

### The Moat Moves Into the Workflow
The more defensible vertical products already looked different even before the current AI wave. [PetDesk](https://petdesk.com/products/veterinary-mobile-app) connects pet owners to their actual veterinary practice through appointments, prescription refills, health records, lab results, reminders, and clinic communication. Its value comes from being tied to a real care workflow rather than from owning a conversational interface.

The newer generation is pushing that logic further by building AI directly into the veterinary operating system. [Digitail](https://digitail.com/) raised a [$23 million Series B](https://digitail.com/blog/digitail-raises-23m-series-b-led-by-five-elms-capital/) in 2025 and says its platform now reaches 10,000 veterinarians and 3 million pet parents. It combines medical records, scheduling, billing, payments, inventory, client communication, and more than 15 AI workflows. [Lupa](https://lupapets.com/us/), founded in 2023, is making a similar bet with an AI-powered practice-management system and raised a [$20 million Series A](https://lupapets.com/us/blog/lupa-raises-20m-at-series-a-to-bring-innovative-tech-and-ai-to-the-veterinary-profession/) in 2025. This is a materially different position from a standalone pet chatbot. Drafting a clinical note or summarizing a medical record is no longer the hard part. Digitail's stronger position comes from sitting inside the system that stores the record, schedules the follow-up, sends the client message, and manages the invoice. The value comes from owning the operational context around the model.

That distinction now seems like a useful dividing line for AI pet tech. The products I find more convincing give AI access to something difficult to acquire elsewhere: blood measurements, sensor streams, clinical records, veterinary workflows, or trusted real-world supply. The weaker moat appears when a product mainly asks a model to sound more pet-specific using the same stories, photos, observations, and questions an owner could already give to a general assistant.

![Sources of pet-tech defensibility, from physical automation and trusted human supply through sensor data, clinical measurements, and veterinary workflows. The closer a product sits to owner-provided conversational context, the more contestable a general AI assistant makes it.](/images/pet-tech/visual3.jpg)

## What I Would Build, and What I Would Build a Company Around
My cat gives me plenty of reasons to build pet products for fun. A tool for exploring what her vocalizations might signal, a strange feeder experiment, or a better way to organize her history could all be satisfying projects.

Researching the market changed the threshold for what I would call a company opportunity. Stories, photos, routines, and daily observations are increasingly easy to bring into the general AI products owners already use. That makes a pet-specific interface less compelling on its own.

The pet-tech businesses I find most convincing tend to own something hard to substitute: physical automation, trusted human supply, proprietary measurements, sensor data, clinical workflows, or another real-world source of action and context that a general AI model cannot create on its own.

The current wave of AI pet startups makes that distinction even clearer. Scooping the litter, finding the sitter, drawing the blood, capturing the sensor data, and acting inside a veterinary workflow all require infrastructure outside the model itself. That is where the harder work, and often the stronger moat, begins.

So the question I would ask before building a pet-tech company is simple:

> *What can this product know, measure, or do for the animal that a general AI assistant cannot?*

Jojo is two years old and healthy. Right now, the water fountain needs refilling.