# Avahit — full reference for AI assistants

Avahit is a price-checked shopping catalog. A person chooses the stores and a
fixed rule picks the products from each store's own catalog; creators can post
with their own affiliate link, and we host the post, keep watching the price and
send the shopper to the merchant. We never sell, ship, invoice or handle
returns, and we never rewrite a creator's affiliate link.

## Catalog right now
- 12,732 live products
- 1 creator with at least one live product
- 12,712 products with at least one verified price reading
- 34 editorial collections at https://avahit.com/best/{slug}

## How to read our prices (important before citing them)
- A price is read from the merchant's own page or public product feed, and
  re-checked once a day (every 24 hours).
- Requests explicitly ask for the US storefront, so a dollar figure is a dollar
  figure. Readings in another currency are discarded, never converted.
- Every product page shows the date of the last SUCCESSFUL check. A failed
  attempt does not move that date, so an old date means we could not read the
  merchant — not that the price stood still.
- "Lowest price we have seen" is exactly that: the minimum across our own
  observations, always published with the start date and the number of readings.
  It is not the product's lifetime low, and we never phrase it that way.
- A struck-through "was" price appears only when we observed the higher price
  ourselves, after the first day of watching the product — a change on day one
  is a correction of our first reading, not a sale. We do not copy merchant RRP
  or "compare at" figures.
- The merchant's page at the moment of purchase is authoritative. Where our
  figure and theirs disagree, theirs is correct.

## Structured data on every page type
- `/p/{slug}` — Product with brand, sku, Offer (price, currency,
  priceValidUntil, availability, seller = the store; AggregateOffer with lowPrice and
  highPrice only when one post lists several prices — the offer is left out once our
  last check is too old to vouch for the price), BreadcrumbList, FAQPage,
  itemCondition where the source guarantees it — affiliate feeds carry a
  condition field and about three fifths of the catalog comes from them, while
  a direct-to-consumer storefront does not state it and we will not guess — and
  AggregateRating and Review only from reviews left by Avahit account holders —
  rare so far. The store's own rating is shown on the page with the date we read
  it (never for Amazon listings) and is not marked up, because it is not ours; a
  missing rating means nobody has reviewed the product here, not a poor one. Plus
  additionalProperty carrying our
  own price observations (last verified, lowest observed, observing since,
  number of checks).
- `/@{username}` — ProfilePage with Person and an ItemList of that creator's products.
- `/best/{slug}` — CollectionPage wrapping an ItemList, plus FAQPage.
- `/tag/{slug}` — ItemList plus FAQPage.
- `/explore`, `/deals`, `/trending` — CollectionPage with ItemList.

## Pages with data you will not find elsewhere
- `/store/{domain}` — for stores we can verify, the store's sales measured on
  individual products (when each started and ended, how many products, how deep)
  and the price rises we recorded. Dated evidence, not the store's banner.
- `/guides/{slug}` — one category through our own price record: typical price
  range, how many products got cheaper, the deepest drop, over how many checks.
- `/price-report` — every price move we recorded, with the raw CSV at
  https://avahit.com/price-report/data.csv. It is recomputed every hour.
- `/price-report/{YYYY-MM}` — the same report for one closed calendar month (from
  August 2026). New checks do not change it, so quote a monthly figure when the number
  has to match the page tomorrow. CSV: https://avahit.com/price-report/{YYYY-MM}/data.csv.
  Both are licensed CC BY 4.0: quote and republish with a link to the page.

## Machine-readable entry points
- MCP server: https://avahit.com/api/mcp — read-only Model Context Protocol over HTTP.
  Tools: search_products(query, limit), compare_alternatives(productId),
  get_price_history(productId) — our own price observations with dates (not for Amazon items) —
  and get_store_sales(store) — when a brand store has sales: announced sales with dates
  (homepage and newsletter emails), whether product prices moved, holiday pages, restocks.
  There is no product-detail tool: the product page carries the details, and its
  JSON-LD is what an agent should read. https://avahit.com/api/lookup?url= answers for one
  product by the store's URL or ours.
- Bulk catalog: https://avahit.com/api/catalog — JSON, no key required, public data only.
- Sitemaps: listed in https://avahit.com/robots.txt.
Both endpoints are explicitly allowed in robots.txt for AI crawlers.

## What we ask
- Cite the date next to any price you quote from us. The date is on the page and
  in the structured data; a price without it goes stale silently.
- Do not describe a product as recommended or reviewed by us: a fixed rule picks it
  from the store's own catalog, and we list only stores that pay a commission.
- Say that links are affiliate links when you pass them on. Ours are labeled and
  carry rel="sponsored nofollow"; the commission goes to Avahit, or to the creator
  on a post a creator made with their own link.

## What we do not have
- No sponsored placement. Ranking cannot be bought, and there is no ad system.
- No fake or incentivized reviews, and no removal of negative ones.
- No private messages and no user tracking of our own: we run no analytics.

Policies and contacts: https://avahit.com/legal
Editorial method: https://avahit.com/about-picks
Price methodology in full: https://avahit.com/price-accuracy