What owners actually go through, before they ask anyone.

Project Pawly starts from one idea: owners don't need a smarter chatbot, they need help that already knows their situation. That only works if you know what the situation usually is. The study is our map of what owners carry: the problems, how long they've lasted, the steps already taken, and what stops people going further. It shapes what the app asks about and what it remembers.

Public discussions from open dog-owner communities, collected through public interfaces. No private messages, no closed groups, no purchased data, and automated moderator posts removed before counting.

Discussions that passed the relevance screen were then read in full by a model and turned into a structured record. We will publish the size of that in-depth set together with its accuracy audit.

• One frozen instruction set. Every discussion was read against the same instructions, locked before the run and never edited mid-study, and returned as a fixed record: the problem, how long it had gone on, what had already been tried, and what stood in the way.
• Checked against the source. Extracted phrases are checked against the original text.
• Only people facing something. Discussions that turned out not to involve an owner with a current problem were counted and set aside.
• Effect sizes, not significance stars. At this scale nearly any difference is statistically significant, so findings are stated as effect sizes with intervals. Free-text answers are grouped into keyword families and the ungrouped remainder is disclosed, so shares are lower bounds.
• What was said, never who said it. No usernames, account identifiers or profile data are kept, and no discussion is quoted. Findings are published only in aggregate.

• It usually isn't a passing annoyance. Most owners describe something ongoing, not a one-off question.
• It has usually been going on a while. By the time people ask strangers, the problem is rarely new.
• They have already tried things. Most owners with a live problem name something they already did about it, and many had already seen a vet or a trainer. What they're after is what that visit didn't settle.
• Money isn't the only barrier. Cost leads, but the next two are not about money: wanting to handle it themselves, and the memory of something that already failed.
• Distress tracks the household, not just the dog. A tight budget is one of the strongest markers of an owner who has run out of ideas.
• It isn't a knowledge gap. As problems get worse, anxiety and guilt rise sharply while diligence stays flat. The same careful owner is there at every level, carrying more.
• In the middle of a problem, people want help. Among what owners say they would pay for, direct help with a behavior problem ranks far ahead of shareable proof like badges or report cards.

• Ongoing, not one-off: a problem stays open as a plan with check-ins, not a single answer.
• Already tried things: it asks what you've tried before it advises, and remembers the answer.
• A failed attempt is a barrier: when a step isn't working, the plan steps back instead of repeating itself.
• The household matters: it asks who's home and what your days look like.
• Not a knowledge gap: one clear step at a time, not more information.

The discussions told us what owners go through. The plans themselves come from approved guide pages, drafted from a separate library of research, veterinary and working-trainer sources, and the food checks match against a product catalog.

Alongside Pawly, the founder studies how people actually use general AI chat, across large sets of public conversations. The pattern that shaped this product: model capability isn't what limits the value people get. A chat box leaves recognizing the problem, supplying context, judging the answer and noticing what happened to the person, and in broad public chat logs few people push back on an answer or check it against what good looks like. Project Pawly is built to carry that work for one job: raising a puppy.

• Composition drives the themes. The largest single community is a pet-bereavement community, which is why grief ranks high. The themes describe this collection, not dog ownership at large.
• Dog owners broadly, not only puppy owners. It is a map of owner problems in general; puppy owners are one part of it.
• People who post, not all owners. Owners who write publicly skew toward those who have a problem and lack an easier place to ask.
• Self-report only. Everything is the owner's own account; no vet record confirms a diagnosis.
• Machine reading is imperfect. Some extracted phrases are lightly rephrased rather than exact, and a small share fail to parse; both are measured per batch.
• Association, not cause. Where a situation goes with higher distress, nothing here shows direction or cause.
• Stated willingness to pay is not spending, and it was measured while problems were live.
• English-language and platform-bound. It reflects the communities collected, not other languages or regions.

Questions about the method are welcome at jake@pawlytechnologies.com.