compare-to-local

Compare to Local (Israel)

Given a product (name, model, or AliExpress listing), find out whether it's available locally in Israel and at what price, by running two separate scoped web searches:

1. <query> site:il — broad sweep across the .il TLD (catches .co.il, .org.il, .net.il, etc. — Ksp, Bug, Ivory, Lastprice, manufacturer pages, smaller shops, marketplaces)
2. <query> site:zap.co.il — Zap is Israel's dominant price-comparison aggregator; this surfaces aggregated listings with multi-vendor pricing

The two searches are run independently and reported separately. Do not collapse them — they answer different questions:

• site:il = "who sells this in Israel?"
• site:zap.co.il = "what's the comparison-shopping price?"

Inputs

• query (required) — the product to look up. Prefer a precise identifier (brand + model number, e.g. Anker 737 PowerCore, Logitech MX Master 3S) over a generic description.
• aliexpress_url (optional) — the AliExpress listing the user is comparing against. If supplied, also extract the AliExpress title and ILS price for the final side-by-side block.
• aliexpress_price_ils (optional) — landed-cost figure if the user has already computed it; otherwise just the listed AE price.

How to query

Use whichever web search tool is available (WebSearch, Tavily, etc.). Fire two distinct searches:

Search 1 — broad .il TLD

<query> site:il

Notes:

• site:il matches any .il second-level (co.il, org.il, net.il, ac.il, gov.il, plain .il). Don't narrow to site:co.il — you'll miss legitimate retailers.
• If the query is an English brand/model, also try a second variant with the Hebrew transliteration if obvious (e.g. Logitech → לוג'יטק). Mark Hebrew variants as a separate sub-search in the report.

Search 2 — Zap aggregator

<query> site:zap.co.il

Notes:

• Zap product pages live at zap.co.il/model.aspx?modelid=... (aggregated model page) and zap.co.il/models.aspx?... (search). The model page is the gold result — it shows a price range across vendors.
• If the top hit is a Zap category page rather than a model page, drill into the model page before reporting prices.

What to extract from each result

For every promising hit (top ~5 per search):

• vendor — retailer name (e.g. Ksp, Bug, Ivory, Lastprice, Zap, manufacturer)
• url — direct product URL
• title — product title as shown on the page
• price_ils — listed price in ILS. If a range, capture both ends.
• in_stock — best-effort: in-stock / out-of-stock / unknown
• notes — any flags worth surfacing (refurbished, parallel import / יבוא מקביל, kit-only, used)

If the search snippet contains the price, you can use it directly. Otherwise fetch the page and parse — but only for the top 2–3 results to avoid wasted fetches.

Israel-specific gotchas

• Parallel import (יבוא מקביל): common on AliExpress-adjacent items. Often cheaper than the official importer (יבואן רשמי) but warranty is shorter and not honoured by the brand's IL service centre. Surface this distinction if visible.
• VAT (מע"מ): prices on Israeli retail sites are normally VAT-inclusive (currently 18% as of 2026; verify if the comparison hinges on it). AliExpress prices to IL are pre-customs/VAT for orders ≥ $75 USD, so always compare landed cost, not sticker.
• Zap "starting from" prices: the headline price on a Zap model page is the cheapest single vendor; click through to see whether that vendor is reputable (Zap shows a star rating). Don't treat the headline price as the realistic floor without verifying the vendor.
• Hebrew vs. English titles: many Israeli retailers list the product in Hebrew but with the English model number embedded. Match on the model number, not the title text.

Output format

Report the two searches separately, then a comparison block if aliexpress_url / aliexpress_price_ils were supplied.

Query: <query>

──────── Search 1: site:il ────────
Top results:
1. <vendor> — ₪<price>   [in stock | OOS | ?]
   <title>
   <url>
   notes: <parallel import / official / refurb / …>
2. …

──────── Search 2: site:zap.co.il ────────
Zap model page: <URL or "no model page found">
Price range on Zap: ₪<low> – ₪<high>   (across <N> vendors)
Cheapest vendor: <name> @ ₪<price>   <url>
Notable vendors: <list>

──────── Comparison (if AliExpress supplied) ────────
AliExpress: ₪<ae_price>   <ae_url>
Cheapest local (site:il): ₪<best_il>   <vendor>
Cheapest on Zap:          ₪<best_zap>  <vendor>
Delta vs AliExpress:      <±%> vs site:il, <±%> vs Zap

Caveats:
- AliExpress price is sticker — add VAT/customs above ₪… threshold for true landed cost.
- Local prices include 18% VAT.
- Warranty: AE = none / seller dispute only; local official importer = 1y+; parallel import = vendor-dependent.

If a search returns nothing useful (genuine zero results, or only irrelevant hits), say so explicitly. Do not invent vendors or prices.

Failure modes / what to do

• No results on either search: report that the product appears not to be sold in Israel under that name; suggest the user try the Hebrew transliteration or the brand-only query.
• Zap returns category page only: report the category URL and note that no aggregated model page exists yet — common for very new or niche products.
• Price visible only after login or in cart: skip silently; Israeli retailers occasionally hide prices for B2B items. Note in the report.
• Captcha / blocked search result: don't loop. Report which source is blocked and continue with the other.

Out of scope

• Does not place orders or check live cart prices.
• Does not evaluate the AliExpress side beyond the inputs given — pair this with evaluate-listing (downstream skill) if landed cost is needed.
• Does not rank vendors by trust/reputation beyond what's visible in the search snippet — that's a separate concern.

Validation checklist (before declaring success)

1. Two separate searches were issued: one with site:il, one with site:zap.co.il.
2. Each search has its own results section in the output (not merged).
3. All prices are in ILS and labelled with ₪.
4. If aliexpress_url was supplied, the comparison block is present and prices are aligned to the same currency.
5. Vendor names and URLs are real (no fabricated hits).