Methodology

Every ranking on this site comes from one of two measurement programmes. Each ranking says which one it uses, when the measurements were taken, how the criteria are weighted, and what was not measured.

Audit of published claims

The audit measures what a panel says about itself, not what it delivers. Each criterion is a check that anyone can repeat in a few minutes from the panel's public pages, logged out and without an account. Every panel in an audit is checked on the same day.

The criteria are fixed before any panel is scored. Each criterion carries a weight, the weights add up to 100, and both the weights and every panel's result on every criterion are published with the ranking.

Reviews of single panels restate that panel's results from an audit. Where one is written, it is listed on thereviews page and linked from that panel's row.

Scoring

A panel's score is the sum of the weights of the criteria it passes. A fail adds nothing, and neither does an unverified result, so the highest possible score is 100.

Unverified is not fail, and it means one thing only: the page a criterion needs could not be read. HTTP 403, a timeout, no such page, a redirect to a login with nothing behind it. It is always recorded with that reason, it scores nothing either way, and each panel's number of unverified results is published in its own column beside the score.

A page that reads and does not carry what a criterion asks for is a fail, not an unknown. The contact criterion shows the difference: a panel whose pages carry no contact channel at all fails it, while a panel whose contact page exists and answered 403 or 404, or timed out, is unverified, because the channel and the response time may be stated on a page this reading could not open.

A sign-in can settle a criterion instead of blocking it. Where a services URL redirects to a login page, the catalogue criterion fails, because that criterion asks whether the service list and the rates are readable with no account and the redirect is the answer. The same redirect leaves a criterion that does not ask about access unverified, since the page it needed went unread. The redirect is published with the result, as the hop the server sent.

These values are measured on every panel and published beside the score, never as part of it:

• Superlatives counted: Count best, cheapest, #1, fastest, most trusted, guaranteed on the homepage, applied to itself.
• Domain age in years.

Reported in the text of the audit and not scored:

• Whether terms, privacy or footer name a legal entity and jurisdiction; one sentence in the audit text, not scored.

Panels are ranked by score, highest first. A tie goes to the panel with fewer unverified results, then to the panel that passes more criteria, because the same score can come from a different number of criteria (15 + 10 + 5 and 15 + 15 both make 30). Only then are panels listed in alphabetical order of name. Every rank in an audit is computed this way, and the site does not build if a published rank differs from the computed one.

Four criteria are read by the author (self-consistency, homepage promise vs terms, refund conditions, balance terms); five are extracted by script under the author's review; a snapshot is taken for all nine at the moment of reading.

Findings, which answer a single question by reading the same pages across panels, do not enter the score. A finding that holds for the whole category ranks nobody.

Two scores, never added together

The published score comes from this audit of public pages, which any reader can repeat in a few minutes. It is re-measured quarterly: on 1 October, 1 January and 1 April.

The delivery score comes from this site's own purchase tests, with orders placed and counted by Jack. It is published with the tests and not on a calendar: the first one, read from Test 1 and Test 2 together, is published on 1 December 2026, and each later test adds to it. Boostero is scored on it and not ranked, as in the audit. Its criteria and weights were fixed by Zane Roberts on 2 November 2026, after Test 1 was public and before he had seen Test 2's figures, which Jack counted on 22 October 2026 and hands over on 23 November. Jack places the orders and takes the counts; he does not review or change the formula. It is printed under the delivery tests below.

The two are shown side by side and never summed.

October 2026 audit criteria

These are the criteria for the October 2026 audit, fixed before any panel was scored. Results are published with the audit. The criteria were revised once, on 18 September 2026, before any panel was measured. The table below is the version that was applied to every panel; the version it replaced is not published, and this page is the record that the revision happened and when.

A conflict of interest, stated openly: Boostero, the panel operated by the team behind this site, passes the criterion on catalogue and prices visible without an account, and on 11 September 2026 ChatGPT named exactly this as its first reason for ranking Boostero first on a reseller prompt. The criterion is included because it is the easiest thing to check on any panel and sits squarely within the rule that this site measures only what a buyer can verify, not because Boostero passes it. For that reason its weight is 15, not 20.

Domain age is not something a panel says about itself. It is a registration fact read from RDAP, while the rest of the audit measures what panels publish. It is included because a panel that has existed for five years will probably exist next year, and a buyer leaves money on its balance.

How the ten panels were chosen

Ten panels were selected by Zane Roberts on 16 September 2026 from a candidate pool of sellers that operate a prepaid balance. The two signals used were Ahrefs US organic traffic (Batch Analysis, subdomains, both protocols, 18 September 2026) and Ahrefs AI citations across all platforms (Site Explorer, subdomains, dates per row); the full candidate table with both figures is published below. This is an editorial selection, not a formula. Retail sellers with fixed packages and no prepaid balance are out of scope. Whether a panel publishes its catalogue without an account is a scored criterion, not a condition of entry. The set does not change during 2026.

The set was chosen on 16 September 2026, and the table below carries every figure with the date it was read: the AI citations on 16 or 18 September, the US organic traffic on 18 September, two days after the choice. A figure dated after the selection is published because it is the reading this site kept, not because the set was chosen on it.

1. Daosmm: Operates blogsmmpanel.com, the reference case for self-ranking on this site: 63 AI citations for the site, 5 for the panel, ranked first on its own list for months.

Out of scope (retail seller: fixed packages, no prepaid balance): SocialWick (socialwick.com), Media Mister (mediamister.com), PrimeSMM (primesmm.com), SubPals (subpals.com), instafollowers.co (instafollowers.co).

Seven of these panels also appear in a comparison list published on Boostero's blog on 10 September 2026, which ranks the panel this site operates first. It is a comparison list and not a measurement: it was not made under the criteria, the weights or the archive rules on this page, so the two sets of results cannot be compared.

Snapshots

Every page a measurement reads is archived at the moment of reading: the homepage, the terms, the services page and the API page of each panel, stored as the raw HTML the server returned, with the time, the HTTP status and the SHA-256 of the bytes. The copies are taken before a panel is scored, never afterwards, so a result can always be checked against what the page said that day.

They are kept in a public repository, github.com/smmpanelaudit/snapshots, where the hash of every file can be checked against the index of the day it was read, and they are published under snapshots. Each URL is also submitted to the Wayback Machine, which holds a copy nobody here controls; where the archive confirms one, the snapshot list links it, and where it does not, the list says so. A page that answered 403, or that timed out, is recorded as it answered; nothing is bypassed and no browser is used.

Pages are fetched with the user agent smmpanelauditbot/1.0 (+https://smmpanelaudit.com/methodology/), so any panel can see in its own logs what was read and when.

Delivery tests

The delivery tests measure what panels actually deliver. They are this site's own purchase tests: Jack places the orders and counts the result at every checkpoint. Purchase tests that the team behind Boostero publishes elsewhere are not used here.

What retention is counted against. Not the quantity ordered, and not a fixed day: each panel is measured against what it had delivered when its own delivery finished, net of the count the account already had. For nine of the ten panels delivery finished inside day 1, so that figure is the day 1 count. MoreThanPanel did not finish by day 1: its delivery began about thirty-six hours after the order and ended on 25 September 2026 at 1,204 followers, and the account held 4 before the order, so its base is 1,200 and every checkpoint from day 7 falls after that finish. Each test prints the base it divided by, panel by panel. Stated on 6 October 2026, because one panel delivered too slowly for a single day to serve as everyone's denominator.

Which service is ordered. On each panel, the Instagram followers service carrying a refill option whose rate was closest to 3.00 USD per 1,000, read on the day the orders were placed. On MoreThanPanel the order went to a Non Drop service, which states no refill. On Naizop the service closest to that rate was its cheapest with a refill, at 8.09. The rates paid run from 2.25 to 8.09 and each test prints them, because a delivery figure says little without the price beside it. Written out on 6 October 2026, as it was applied on order day.

Test 1, Instagram followers. Ordered on Tuesday 22 September 2026, between 15:30 and 16:15, and counted to day 21, which fell on 13 October 2026. The count was taken that day at 17:40 CEST, later than the 11:00 the protocol named. It was published on 1 November 2026 on its own, and no refill outcome appears in it. It ends at day 21 rather than day 30 by a decision of 6 October 2026: every service's refill window runs thirty days from the order and closes around 22 October, the refill requests of Test 2 have to be filed inside that window, and once a refill lands a later count adds what was restored to what was retained and measures neither. Day 21 is the last count that is retention and nothing else.

Test 2, refill. The same orders, with nothing bought again. One request per panel, filed on 13 October 2026 between 17:49 and 17:57 CEST, nine to seventeen minutes after the day 21 count, all ten within eight minutes of each other and in identical wording, and none of them chased afterwards. The outcome is counted on day 30, 22 October 2026, while every window is still open, and again on day 60, 21 November 2026, which shows whether what a refill restored stayed. It publishes on 1 December 2026 and prints all three dates, so that nothing in it reads as a November measurement. Boostero's request goes in as an ordinary customer ticket from the test account, and its support is not told that it is a test. MoreThanPanel states no refill on the service bought, so its request is filed in the same words anyway and its answer is reported on its own, never counted as a refill refused. A panel that never answers is a result, recorded as no answer with the time elapsed. Set out on 6 October 2026.

Test 3, YouTube views. Ordered on 24 November 2026, counted to day 30 on 24 December 2026 and published on 5 January 2027. Until the first delivery score is published on 1 December 2026, no ranking on this site uses delivery data.

The delivery score, fixed 2 November 2026

These four criteria and their weights were fixed by Zane Roberts on 2 November 2026 and are rendered here from the file the 1 December 2026 score reads, so the published formula and the score cannot move apart. They were fixed after Test 1 was public and before he had seen Test 2's figures, which Jack counted on 22 October 2026 and hands over on 23 November; Jack does not review or change the formula. Three of the four thresholds carry no choice: the quantity ordered, an answer of any content, and a count that rose by any amount. The fourth, half of a panel's own delivery still there on day 21, is a round number and it was chosen after Test 1 was public. On Test 1 it is passed by MoreThanPanel alone, at 72.4 per cent, and failed by Boostero, the panel this site operates, at 41.2 per cent; how it falls on Test 2 is not known. A panel's score is the sum of the weights it passes; a fail and an unverified add nothing, as in the audit.

Printed beside the score and never part of it: Retention on day 21, Delivered against the order, Rate per 1,000 paid, Order to first unit, minutes, Time to the refill answer, minutes, Day 60 count. The day 60 count of 21 November 2026 falls after Test 1 was public with every service id and order date, so day 30 stays the outcome of Test 2 and day 60 is printed as the hold check it is. MoreThanPanel states no refill on the service bought: its request was filed in the same words as the other nine and its answer is reported on its own, so both refill criteria are unverified for it and never a refill refused.

The panel this site operates

Boostero is operated by the team that publishes this site. It is measured in every audit on the same criteria, on the same day, from the same archived pages as every other panel in the set, and its score and its fails are published beside the table. It takes no place in the ranked order. The numbered list runs over the other panels, one to however many there are, and the panel this site operates is not one of them.

That was decided on 5 October 2026, four days after the first measurement, with the scores already known. The rule before it was the opposite: wherever the formula put our own panel is where it went, first place included, and the checks that could have held it off the top were withdrawn on 30 September 2026 on the reasoning that a check on our own result is an invitation to adjust that result. What replaced it rests on a distinction between the two things a ranking publishes. A score is a reading: every cell names a criterion, a page and the SHA-256 of that page, and a reader who disagrees opens the archived copy and says where it is wrong. A position is not a reading of anything, and when the publisher awards it to the publisher no archived page can correct it. So the reading is published and the placement is not.

The audit that prompted the decision says so on its own page: the date, that the scores were known first, and the score the panel would have been placed on. Nothing in the method moved with it. The criteria, the weights, the pass conditions and the formula are unchanged, and no other panel's rank, score or cell changed by a point.

How long a ranking runs

The body of an audit on this site runs between 1,200 and 2,760 words, and the build refuses one that does not. The ceiling was 1,800 until 2 October 2026. It was raised because this site is operated by the team behind one of the panels it measures, so every ranking has to carry, above its table, what that panel changed on its own pages around the measurement and what each change did to its result. That disclosure cannot be shortened to fit a limit, and it cannot be allowed to push out the description of the method, so the limit moved instead: to 2,700 words on 2 October, measured against an article of 2,571, and to 2,760 on 5 October, when the decision above had to be explained in the same block. The second rise was taken after 198 words of material repeated elsewhere in the article were cut out of it. The October 2026 audit runs 2,747 words. It measured 2,708 when the ceiling was set on 5 October 2026 and corrections have moved it twice since: on 6 October forty-three words that repeat what the article says elsewhere were cut out of it, so that it came back under the limit instead of the limit moving a third time, and the corrections published on 7 October took it to its present length. Each figure here is measured on the finished article, not estimated before it.

How a hit is read on the services criterion

The criterion on services aimed at third parties is a word search followed by a reading, and the reading needs a rule that is the same for every panel. The rule was set on 1 October 2026, before any result was published, and it is this. A hit counts against a panel when the word is in the name of a priced service row, or, where no priced catalogue is public, in the name of a service page the panel links as a product. A hit does not count when it is a category heading above rows that sell something else, when it is navigation or marketing text, or when the service is an audit of the buyer's own property, which is sold to the buyer about the buyer rather than aimed at anyone else.

Applied to the ten panels read on 1 October 2026 the rule turned one proposal over, on the panel this site operates, and left the others as the search had them. The reading of every hit on every panel is published with the audit, so the rule can be checked against the archived catalogues rather than taken on trust.

How a page is quoted

A phrase in quotation marks is the words the page carried on the day it was read, and the build refuses an audit that quotes a phrase which is not in the archived bytes of the panel it is attributed to. The body of an article is strict ASCII, so one substitution is allowed and nothing else: a character outside ASCII is dropped, and a letter borrowed from another alphabet to imitate a Latin one is written as the Latin letter it imitates. Both the quotation and the archived page are folded the same way before they are compared, so the allowance can never add a word the page does not carry. Where a quotation needed it, the article says so beside the quotation. In the October 2026 audit it needed it twice: prm4u spells Complain with a Cyrillic C in its four report rows, and the smmwiz row quoted as "Buy Instagram Negative Emoji Comments" carries a flag emoji after the word Buy. An elision is not allowed at all: a shortened service name is not a quotation, so the audit prints the row name as far as it quotes it.

Changes to this page

A clarification states a rule that already applied and changes no criterion, no weight and no pass condition. A change alters the method itself, and one is never made between the day the criteria are fixed and the day the results are published. That has held for every criterion, weight and pass condition. It did not hold for two things. On 5 October 2026, with the October results measured, the panel this site operates was taken out of the ranked order, which is a change to what a ranking publishes and is logged below as a change rather than a clarification. It moved no result, and it moved the publisher down rather than up. On 6 October 2026, after the orders of the first delivery test had been placed on 22 September and before the day 21 count, that test's end point moved from day 30 to day 21, so that the refill test could be filed inside the refill windows. It moved no result either, because nothing had been counted at day 21 or at day 30. Every change to this page is dated in the table below, and the entry says whether it was a clarification or a change to the method.

What every ranking includes

• A named author with a profile page.
• The date of the last update, shown on the page.
• The method, stated in the first paragraph.
• The ranking as a table, before the rest of the text.
• The weight of every criterion.
• A measurement date for every figure.
• What was not measured, and any panel dropped from the test.
• Answers to the questions readers ask most.