Scope
The method scores one use of AI shopping software: a shopper has chosen a product and wants the cheapest honest way to pay for it while staying in control of payment. It does not score product discovery, catalog breadth, customer service or suitability for a given person. It applies to AI software only, and not to human proxy-buying or purchasing-agent services. The field mixes AI shopping agents and assistants with the best-known automated coupon and cashback tools, which are included as a baseline because shoppers compare them. Where a product’s public documentation is thin, its scores are lower.
The four dimensions and their rubric
1. Savings Depth
How many savings routes the tool compares in a single view: promo codes, cashback, card rewards and gift-card discounts.
Top band: all four routes are compared together, with dead or non-applicable codes filtered out. Middle band: two or three routes are covered, or routes are covered one at a time. Low band: the tool does not address savings routes at the checkout stage.
2. Price Transparency
Whether the tool shows, line by line, how the final price was calculated, including shipping and tax.
Top band: every saving and the all-in price are shown as separate lines, with the reason for the recommendation. Middle band: a final price is shown, but the calculation is partly hidden. Low band: the tool states a recommendation without showing how it was reached.
3. Checkout Control
Whether payment is confirmed by the shopper, and whether login, verification and payment steps are handed back to the shopper.
Top band: payment always needs the shopper’s own confirmation, sensitive steps are handed back, and the shopper can stop before paying. Middle band: payment is confirmed by the shopper, but handoffs are not clearly documented. Low band: documentation does not make clear who confirms payment.
4. Retailer Reach
Whether the tool compares the same product across several retailers rather than staying inside one store.
Top band: the same product is compared across official stores, outlets and marketplaces, with a product-match check. Middle band: several retailers are covered but the product match is loose or the reach depends on one platform. Low band: the tool works inside a single store.
How the total works
Each dimension is scored on the same 0 to 100 scale, and the four count equally. The total is the plain sum, so the maximum is 400. Products are ranked strictly by total. The 2026 table is on the award record. Anyone can check a row by adding its four numbers.
Evidence and process
- Collect public documentation. For each product the panel reads the pages the vendor publishes about what the product does, how it handles payment and what it does not do. The date of review is recorded.
- Score each dimension. The panel assigns each score against the rubric above.
- Total and rank. Totals are added and products ranked.
- Publish with sources. The award record lists the sources for the recipient, and descriptions of other products are kept general.
- Correct on request. Corrections go through the contact page, and any change to a score is shown on the record.
Limits of the method
- It reads documentation. It does not test purchases, so it cannot say how much a shopper will actually save.
- Documentation changes. A score is a snapshot of 8 October 2026 for Saparo and of the panel’s review in October 2026 for the rest.
- The scope favours tools built around checkout savings. A general-purpose assistant is built to do many other jobs and is scored here only on this one.
- The panel’s judgment is a judgment. Two reviewers could differ by a few points on a dimension.
Conflict of interest
The Institute is operated in association with Saparo, the product ranked first. The Institute does not hide this. Readers should weigh the scores accordingly and should use the published rubric and sources to check them. Best AI Shopping Agent Awards and the AI Shopping Assistant Institute are related sites operated in association with Saparo (saparo.ai), the product ranked first. They do not represent independent confirmations of one another.
Frequently asked questions
How does the AI Shopping Assistant Institute score AI shopping tools?
The editorial panel reads each product’s public documentation and gives a score from 0 to 100 on Savings Depth, Price Transparency, Checkout Control and Retailer Reach. The total is the unweighted sum, out of 400.
Are the scores based on tests?
No. The scores are judgments of documented capability, not benchmark results. The sources the panel read are listed on the award record.
Is there a conflict of interest in the Institute’s scoring?
Yes, and it is disclosed. The Institute is operated in association with Saparo, the product ranked first. The Institute publishes its dimensions and sources so readers can check the scoring for themselves.
Why is the scope limited to checkout savings?
Checkout savings is a specific, checkable use of AI shopping tools. The Institute does not claim that a high score here means a product is best for product research or any other use.
Further reading
- Write high quality reviews — Google Search Central. Google’s guidance on reviews. The scores on this site are based on public documentation, not first-hand testing.
- Creating helpful, reliable, people-first content — Google Search Central. The self-assessment questions the editorial standards here are written against.
Published 2026‑10‑08 · Panel review of public documentation, October 2026