AIPlayPark

Methodology

How AIPlayPark Ranks AI Vendors: Two Rubrics, Sources, Bias and Corrections

Every rubric-scored ranking on AIPlayPark is a sort of one table. The healthcare table has 39 vendors, eight criteria and 12 points; the voice agents table has 28 vendors, eleven criteria and 12 points. Every answer is taken from the vendor's own published pages and quoted with a link and a date. This page is the whole method for both, so anyone can rebuild any list by hand.

/
By the AIPlayPark editorial team, part of MedReception AI

Who chose these criteria, and the bias to weigh

The eight healthcare criteria, the eleven voice agent criteria and their point values were chosen by MedReception AI's team, and MedReception AI is one of the 39 vendors scored on the healthcare rubric. That is a bias, and readers should weigh it: a rubric written by one vendor can favor the things that vendor does. What limits the bias is that nothing on these pages rests on our opinion. Every criterion is answered from each vendor's own published pages, the quote and its source are shown next to the value, and the sort is arithmetic. A competitor that scores higher ranks higher. If you think a criterion is wrong or missing, or a quote is out of date, the corrections section below says how to tell us.

The healthcare rubric (12 points)

Criteria, how each is scored, and the maximum points
CriterionHow it is scoredMax
HIPAA complianceStated by the vendor on its own pages: 1 point. Verified by a named third party (for example Compliancy Group): 2 points. Not stated: 0.2
Business associate agreement (BAA)The vendor states it signs, provides or executes a BAA: 2 points. States it does not, or does not state: 0.2
CoverageStates 24/7 (around the clock) answering: 1 point. Business hours only, or not stated: 0.1
Books appointmentsStates it books or schedules appointments into a calendar or EHR: 2 points. Takes messages only, or not stated: 0.2
Named EHR integrationsCounts only EHR or practice management systems the vendor names on its pages. 1 to 4 named: 1 point. 5 or more named: 2 points. None named, or a count claim with no names: 0.2
Published pricingA price with a dollar (or other currency) figure is published: 1 point. Quote only, or not stated: 0.1
Healthcare focusThe pages state the product is built for medical, dental or healthcare practices: 1 point. Serves many industries, or not stated: 0.1
LanguagesStates support for more than English: 1 point. English only, or not stated: 0.1

Two more fields are recorded but not scored, because they describe a choice rather than a strength: whether an AI, live agents or a mix answers the phone (used to split the AI receptionist lists from the answering service lists), and the lowest published monthly price (shown for context; never computed from per-minute rates).

Healthcare tie-breaks, in order

  1. a BAA stated on the vendor's pages
  2. published pricing
  3. more named EHR integrations
  4. vendor name, A to Z

Example: DoctorConnect and VoxyHealth both score 10 of 12 and both state a BAA and quote-only pricing; DoctorConnect names 10 EHR systems to VoxyHealth's 7, so it ranks first.

The voice agents rubric (12 points)

The second category, “AI voice agents and AI phone answering for any business”, scores 28 vendors on eleven criteria read from their own pages on 2026-11-05. The publisher's own product, MedReception AI, is not in this category: it serves medical and dental practices only, so it is scored on the healthcare rubric above and appears in no voice agents list. The value definitions below are the ones in our published data file, word for word.

Voice agents criteria, how each is scored, and the maximum points
CriterionValues and pointsMax
Published pricingpublished: a dollar or euro figure appears on the vendor's pages (1 point). quote: the vendor states contact sales, custom pricing or book a demo and publishes no figure (0 points). unknown: not stated (0 points).1
Free trial or free tierfree: a free trial or free tier is stated (1 point). none-stated: pricing pages read and no free trial or free tier stated (0 points). unknown: not stated (0 points).1
Answers inbound callsyes: answers inbound calls (1 point). unknown: not stated (0 points).1
Places outbound callsyes: places outbound calls (1 point). no: the vendor states the agent does not place calls (0 points). unknown: not stated (0 points).1
Books appointmentsyes: books appointments or writes to a calendar (1 point). unknown: not stated (0 points).1
Named integrationslist of NAMED integrations exactly as the vendor names them (CRMs, calendars, phone systems, helpdesks, Zapier/Make, EHRs); a bare count claim is a fact but not a name. none or unknown: 0 points. 1 to 4 named: 1 point. 5 or more named: 2 points.2
Handoff to a personyes: transfers or escalates to a human (1 point). unknown: not stated (0 points).1
Languagesmultilingual: more than English, count in `count` when stated (1 point). english: English only (0 points). unknown: not stated (0 points).1
Coverage24/7: around-the-clock answering stated (1 point). unknown: not stated (0 points).1
Security statementstated: names SOC 2, HIPAA, GDPR, ISO 27001 or PCI as something it holds or complies with (never 'verified') (1 point). unknown: not stated (0 points).1
Channelsvoice-plus: voice AND at least one of SMS, chat, email stated (1 point). voice: voice only stated (0 points). unknown: not stated (0 points).1

Voice agents tie-breaks, in order

  1. published pricing
  2. more named integrations
  3. a security statement on the vendor's pages
  4. vendor name, A to Z

Ten vendors score 12 of 12, so these tie-breaks decide the top of every voice agents list. Each list page says so where it applies, and no other separation is invented.

  • The unknown rule. If no fact on the vendor's own pages supports a value the value is 'unknown' (never a guess), evidence null, score 0; page copy says 'not stated on the pages we read'.
  • Evidence. Every non-unknown value carries evidence = {text: a fact written from the vendor's own page, source_url: the page it was read on, as_of: the date read}. The same text appears verbatim in facts[].
  • Conflicts. When two facts conflict the more conservative value is chosen and note explains. The note is shown on the vendor's profile as its scoring note.
  • Not scored, shown for context. The kind of product (a finished service you turn on, a platform you build an agent with, or a human plus AI service), which splits the lists, and the starting price: lowest published PAID price expressed per month, as the page displays it (annual-billing per-month figures count and the billing term is in the evidence); a $0 free tier is recorded under trial, not here; never computed from per-minute, per-call or per-credit rates.

How the facts are collected

We read each vendor's own website on the date shown next to every quote (the healthcare table mostly on Nov 2 and Nov 4, 2026; the voice agents table on Nov 5, 2026) and record what it states, near-verbatim, with the page URL. A vendor's pages are the only source for anything we say about that vendor. We do not infer HIPAA compliance, a BAA, pricing or an integration from reviews, press, badges without text, or our own experience.

  • Not stated is a value. If no page we read addresses a criterion, the value is “not stated on the pages we read” and the score is 0. It describes the pages, not the product.
  • Conflicts resolve conservatively. When two pages disagree, the lower-scoring value is used and both quotes are kept. Goodcall's site says HIPAA compliant on one page and makes no HIPAA statement on its pricing page, so its HIPAA value is not stated; both quotes appear on its profile.
  • A negative statement is shown as the vendor wrote it. Smith.ai's FAQ states it is not HIPAA compliant and cannot handle calls involving PHI. The rubric has no negative value, so it scores 0, and every page that shows Smith.ai says what it states rather than “not stated”.
  • Count claims are not names. “More than 80 integrations” scores only for the systems the page names. The claim is quoted next to the score so you see both.
  • Certifications are not verification. HITRUST, SOC 2 and similar are security frameworks; they do not move the HIPAA value from stated to verified.

How MedReception AI is scored

MedReception AI is scored on the same eight criteria from its published facts: HIPAA compliance verified by Compliancy Group with a BAA included; 24/7 answering and appointment booking; 19 live EHR integrations, each named; Studio at $299 per month for 500 minutes with managed plans from $495; a medical and dental focus (veterinary practices are not served); and 59 languages. Those facts give it 12 of 12. We publish no outcome statistics, testimonials or review counts for it, and none for any other vendor.

How each list is built

A list is an inclusion rule applied to the table, then the sort above. The rule is printed on every list page. Current lists:

Healthcare and medical office AI (16 lists)

Voice agents and phone answering for any business (14 lists)

A list in this category publishes only when at least 4 vendors meet its rule; a smaller result is dropped, never padded.

Request a correction

If a quote is out of date, a value is wrong, or your pages now state something they did not on the date shown, tell us through the contact page. Include the URL of the page on your own site that states the fact. We re-read that page, update the quote, the value and the date, re-sort every list the vendor appears on, and change the Updated date on each affected page. We do not accept facts from any other source, and no vendor can pay to change a score.

Related sites

/
By the AIPlayPark editorial team, part of MedReception AI