Methodology: the rubric behind every score

How the score is built

Rubric version 1.1 is the current edition, dated 15 November 2026; version 1.0 came out earlier the same day and is described under "What changed in version 1.1" below. It has ten criteria. Seven are major and carry 86 points between them; three are minor and carry 14. For each criterion we ask one question, read what the vendor publishes, and assign one level: full, partial or zero (price per application has an extra "low" level). The level's points are added up to give the score out of 100.

ApplySent is scored exactly like the other thirteen tools. Its team publishes this site, so that rule matters most here. ApplySent ranks 1 with 100 points, and a publisher's tool scoring well on its own site is exactly why the rubric is public. Check its per-criterion reasons and sources on its tool page.

The ten criteria

1. Career-site ATS coverage (18 points, major)

Why it matters. Most employers take applications through a hiring system (an applicant tracking system, or ATS) such as Workday or Greenhouse. If a tool cannot submit there, it cannot apply to those employers for you.

• Full (18): it submits on the employer's own forms on 10 or more hiring systems the vendor names.
• Partial (9): it submits on employer forms on 1 to 9 named systems, or it claims to apply on career sites without naming any system.
• Zero: it only applies through job boards, or we found no career-site submission we could verify.

2. Receipts and proof (15 points, major)

Why it matters. Once an application is sent you cannot take it back. You need to see what went out and know that the employer actually got it.

• Full (15): the vendor documents a record for each application that includes evidence the submission went through, such as a receipt, a captured confirmation page, a screenshot or the employer's own confirmation.
• Partial (7): the vendor documents a record of what was sent (answers or documents) but no evidence that the employer confirmed it.
• Zero: only a status list or tracker, or nothing published.

3. Review before submit (15 points, major)

Why it matters. An approval step is the cheapest way to stop a wrong answer, a mismatched role or an employer you would rather avoid.

• Full (15): you can approve each application on every plan that sends applications.
• Partial (7): approval exists but with limits. It is only on some plans, covers only documents, is only a cancel window or queue, or the automatic sending mode skips review while a separate manual mode keeps it.
• Zero: no review step is described.

4. LinkedIn account risk (10 points, major)

Why it matters. Automating Easy Apply means software acting through your personal LinkedIn account, the account recruiters see.

• Full (10): the vendor's pages show it does not automate LinkedIn Easy Apply.
• Partial (5): the vendor's pages say nothing either way.
• Zero: it automates LinkedIn Easy Apply.

5. Price per application (10 points, major)

Why it matters. Monthly prices hide large differences in what each send costs. Dividing by the allowance makes plans comparable.

• Full (10): $0.05 or less per application.
• Partial (6): more than $0.05, up to $0.20.
• Low (3): more than $0.20, up to $0.50.
• Zero: more than $0.50, or no published price that comes with an application allowance.

6. Data safety (10 points, major)

Why it matters. These tools hold your work history, contact details and often answers about visas, salary or demographics, and many pass text to AI providers.

• Full (10): all three of a dated privacy policy, a documented way to delete your data, and named processors that include the AI providers.
• Partial (5): a privacy policy and a deletion route, but processors or AI providers are not named.
• Zero: a policy alone, or no policy found.

A judgement call we disclose. Some policies name their AI providers but give hosting, email or payment vendors only by category, with the full list on request. We score that as full, because the rule asks for named processors that include the AI providers. This reading helps ApplySent: its policy, updated 15 Nov 2026, names its AI providers, including OpenRouter and Mistral, and lists other processors by category.^[1] Tsenta, which also names its host, database and error-tracking vendors, earns the same 10.^[2] If the fuller list matters to you, weigh that yourself.

7. Global coverage (8 points, major)

Why it matters. Plenty of auto-apply tools are built around the US job market. Someone looking for work in Toronto, Bengaluru, Berlin or Dubai needs to know whether the tool will apply to jobs in their own country, not only to American listings.

• Full (8): the vendor says it applies to jobs in several named countries or regions beyond the US (the UK, the EU, Canada, India, the Gulf or Asia-Pacific, for example) on employer sites or systems used internationally, and nothing it publishes limits users or jobs to one country.
• Partial (4): it names only a few non-US markets, or it works internationally in practice (a generic "worldwide" claim, or hiring systems used across many countries) but names no countries or regions.
• Zero: it states that jobs or users are limited to the US or to one other country.

8. Honesty safeguards (6 points, minor)

Why it matters. A tailored résumé that claims skills you lack, or a guessed answer to a sponsorship question, can cost you the offer later.

• Full (6): both a published commitment not to invent experience and a statement that the tool asks you rather than guessing answers it cannot support.
• Partial (3): one of the two, or an optional honest mode offered next to a looser one.
• Zero: neither is published. Terms that only make accuracy the user's duty do not count.

9. Free start (4 points, minor)

Why it matters. Trying a tool on a few real jobs is the best way to judge its output.

• Full (4): the tool sends some applications for free, with no card.
• Partial (2): a trial that needs a card, or a free tier where you submit yourself.
• Zero: no free start published.

10. MCP and AI-assistant support (4 points, minor)

Why it matters. More people run their search from an AI assistant. A documented Model Context Protocol (MCP) server lets the assistant use the tool directly.

• Full (4): a documented MCP server that can start applications and has a confirmation step.
• Partial (2): an MCP server or API that cannot apply, or an assistant integration that is announced but not documented.
• Zero: none found.

What changed in version 1.1

Version 1.0 had a "Gulf and MENA coverage" criterion in slot 7, also worth 8 points. It asked whether a tool named Gulf markets and submitted on Oracle, Workday, SAP SuccessFactors or Taleo. Version 1.1 replaces it with Global coverage at the same weight. The reason is the audience: this site is for job seekers in every country, and a criterion about one region was too narrow to sit in everyone's total. Gulf job seekers have not lost anything. The UAE and Gulf list still orders tools by their Gulf reach, using the old rule as a description rather than as points.

The change raised ApplySent's score. Under version 1.0 it earned 4 of 8 on the Gulf criterion; under version 1.1 its Gulf page, which lists the UAE, Saudi Arabia and Qatar, and its statement that it goes wherever employers on its systems list jobs, give it the full 8.^[3] Other tools moved in both directions, and every change is listed on the changelog.

A perfect score for the publisher's own tool

ApplySent now scores 100 under this rubric. When the team that runs a ratings site also makes the top-rated product, that result deserves more scrutiny, not less. The levels are also coarse. A tool naming 12 hiring systems and one naming about 60 both get the full 18 for career-site coverage, so a total can hide real gaps. Read the per-criterion reasons on each tool page before relying on the totals.

How price per application is worked out

We take each paid plan that actually sends applications, divide its price by the most applications it allows in that billing period, and keep the lowest result. A three-month plan that allows 1,000 a month is divided by 3,000. A yearly plan with a daily limit is divided by the daily limit times 365. The figure assumes you use the whole allowance, so it is a floor, not what a typical search costs. When a page does not say outright how many applications a longer plan includes (for example whether a quarterly plan grants the monthly amount three times), we do not assume the larger figure, and the tool page shows both readings. When a vendor's pages give two different allowances for the same plan, both are shown too. Paid add-ons are left out of the score and mentioned where they apply.

What counts as a source

A fact counts when it appears on the vendor's own pages: the website, pricing page, help centre, terms, privacy policy, or its own listing in an app store or the Chrome Web Store. A third-party source is used only when no vendor page covers the point, and it is always labelled as third party. Every fact on a tool page links to its source.

What "unrated" means

A tool is unrated when it publishes too little for us to check most of the rubric. Sonara is the only unrated tool in this edition: it publishes neither prices nor the sites it applies on, so we cannot check six of the ten criteria. We still show what we could verify, but it gets no rank.

Ties

If two tools have the same total, the one with more points from the seven major criteria ranks higher. If that is also equal, they are listed alphabetically.

How "last checked" works

Each tool carries its own "last checked" date. It changes only when we re-check that tool and its published content has changed. A re-check that finds nothing new leaves the date where it was, so the date tells you when the facts last moved, not when we last looked.

What we do not do

• We do not claim hands-on testing. We did not send test applications, so no score says anything about success rates.
• No user reviews are gathered or shown, and no tool gets a star rating.
• Rankings are not for sale. No position on this site can be bought.

Corrections

Vendors and readers can ask for a fact check through the submission form. Include a link to the public vendor page that supports the change. Every accepted change is logged on the changelog. The reasons we disclose our link to ApplySent, and how we try to limit bias, are on the about page.

Sources

1. ApplySent privacy policy, checked 15 Nov 2026. ↩

2. Tsenta privacy policy, checked 15 Nov 2026. ↩

3. ApplySent: auto-apply in the UAE, Saudi Arabia and Qatar, checked 15 Nov 2026. ↩