DogBreedCompass

Which human foods dogs can have, and which parts of them cause the trouble

Every verdict on these 258 pages comes from a named, fetchable veterinary or kennel-club source, and the source is printed on the page it supports. Where a food is fine in one form and dangerous in another — corn kernels against the cob, banana flesh against the peel, ripe tomato against the plant — the page says which is which, because that distinction is usually the whole answer.

What you will not find is a portion table. The file deliberately carries no per-dog portion table. The defensible general rule is the WSAVA ceiling on treat calories; food density, body condition, current diet, and disease history all change the real answer, and no source in this set publishes a per-breed or per-weight count of strawberries, apple slices or shrimp.

If your dog has eaten something now

Do not wait for a web page, and do not induce vomiting or give a home antidote unless a veterinary professional tells you to.

  • ASPCA Animal Poison Control Center: (888) 426-4435 — 24/7, 365 days a year. A one-time consultation fee applies; the ASPCA does not publish the amount on this page.

vegetable

fruit

meat/seafood

processed/composite

grain/starch

herb/spice

dairy

dessert/snack

condiment/sauce

prepared meal

nuts/seeds

beverage

vegetable/seasoning

dessert/candy

meat

seafood

animal product

fruit part

nuts

fruit/vegetable

fungus

nuts/spread

protein

beverage/environment

dough

grain/vegetable

oil/fat

plant part

prepared vegetable

processed food

seasoning

spoiled food

sweetener

vegetable part

How much of anything is too much

All treats combined should generally stay within 10% of daily calories for a healthy adult dog in ideal body condition; individual needs vary.

Average healthy adult dog in ideal body condition. A veterinary team should set the individual dog's daily energy requirement. The reference allowance is printed in full on each food page, and the separate question of how many calories your dog needs in total is handled by the dog food calculator.

Source: WSAVA Global Nutrition Committee, Feeding Treats to Your Dog. Retrieved 2026-09-07.

What this index does not answer yet

This is batch one: 258 foods out of an inventory of 391. The gap is not laziness, it is the sourcing rule. A food gets a page when a named veterinary or kennel-club reference has ruled on it; until then there is no page, because the alternative is guessing a safety verdict from a food that looks similar, and on a pet-health page that is how readers get hurt.

Foods with heavy search demand and no sourced verdict here yet, in demand order: peanuts, pistachios, cinnamon, pickles, cashews, honey, pork, ham, bacon, olives, cabbage, cauliflower. Their absence is not a verdict in either direction. Until one of them appears on this index with a source attached, ask a veterinary team rather than reading anything into the silence.

Sourcing and review

Dataset 2026.09, last verified 2026-09-07. Selection method: Intersection of measured US Semrush search volume (KEYWORDS_SEMRUSH_REAL_2026-09-07.tsv, 7,618 keywords) with rows in B1_Can_Dogs_Eat_MASTER.csv marked publish_ready and carrying at least one source URL. Foods with high volume and no sourced row were skipped rather than filled in from general knowledge. Batch 2 (2026-09-07, +39 foods): highest-volume research_required rows sourced via a two-stage workflow - one agent fetches an authoritative veterinary/kennel-club page and quotes it verbatim, a second, independent agent re-fetches every cited URL and confirms the quote is genuinely on the page before publish. 39 confirmed, 0 rejected by the adversarial check, 1 (black pepper) left unpublished because no authoritative source was found. Batch 3 (2026-09-07, +45 foods): next tranche by measured volume, same two-stage source+verify workflow. 50 attempted, 45 confirmed, 0 rejected by the adversarial check, 5 (sesame seeds, jicama, guava, oysters, paprika) left unresolved for lack of an authoritative source. Batch 4 (2026-09-07, +15 foods): the highest-volume rows that were ALREADY publish_ready with sources in the master CSV - no sourcing workflow needed, only the page prose written from the facts already in the row. Two earlier batches spent ~14M subagent tokens sourcing research_required rows at 5,400/mo and below while these, at 18,100-33,100/mo, sat unused. Batch 5 (2026-09-07, +73 foods): the remainder of the already-sourced set, same one-stage write. Each agent read its own row out of the master CSV rather than being handed the facts inline, which kept the arg payload small enough to pass the real slug list instead of a retyped one. Batch 5 (2026-09-07, +11 foods): the last publish_ready rows in the master CSV that had measurable distinct demand and no live sibling already answering the query. The other 16 publish_ready rows were deliberately left unpublished - each is a near-duplicate of a page already live (baker chocolate, chocolate-covered raisins, peanut butter with xylitol, peach pit, sugar-free candy/gum/peanut butter, macadamia nut cookies, onion soup mix, fruit cake, zante currants, apple pie, wine, liquor, energy drinks, compost) with no measured search volume of its own, and publishing them would add thin siblings rather than answers. Batch 6 (2026-09-07, +27 foods): the first tranche of research_required rows sourced and independently re-verified via a 2-stage agent pipeline (source+quote, then adversarial re-fetch of every cited URL). Prose was written from the verified facts only, with a second adversarial pass checking for fabricated sources/scales/doses before integration. The run hit an account spend limit partway through 52 foods; this is the 27 that completed both stages before the limit hit. The remaining 25 (and a separate 18-food retry batch) are queued for the next pass. Batch 6 (2026-09-07, +25 foods): the first tranche of research_required rows sourced and independently re-verified via a 2-stage agent pipeline (source+quote, then adversarial re-fetch of every cited URL). Prose was written from the verified facts only, with a second adversarial pass checking for fabricated sources/scales/doses before integration. The run hit an account spend limit partway through 52 foods; this is the 27 that completed both stages before the limit hit. The remaining 25 (and a separate 18-food retry batch) are queued for the next pass. Batch 6 (2026-09-07, +3 foods): the first tranche of research_required rows sourced and independently re-verified via a 2-stage agent pipeline (source+quote, then adversarial re-fetch of every cited URL). Prose was written from the verified facts only, with a second adversarial pass checking for fabricated sources/scales/doses before integration. The run hit an account spend limit partway through 52 foods; this is the 27 that completed both stages before the limit hit. The remaining 25 (and a separate 18-food retry batch) are queued for the next pass. The editorial policy sets out how DogBreedCompass sources and corrects this material, and every reference is listed on the sources page.

These pages are reference material, not veterinary advice. A dog with a medical condition, a prescription diet or a suspected ingestion needs a veterinarian or a poison control centre, not a web page.

We research and check what we publish, but we do not warrant that any figure, date, price, or statement is accurate, complete, or current, and we accept no liability for any loss, cost, injury, or other harm arising from the use of this site. Use it as one reference among several. You remain responsible for your own decisions and for your dog's care - confirm anything that matters with a licensed veterinarian or another qualified professional before you act on it. Full disclaimer.