Apollo.io

Fails the removal test, and Apollo supplies the proof itself: the basic tier at $49 per user per month ships with no AI features at all and is sold as a complete working product. A vendor that packages and prices a fully functional plan with the AI removed has answered this axis in its own price list. The substrate is a contact database, a sequencer and a dialer; the AI Assistant arrived March 2026 and the AI power-ups are tier-gated capabilities layered on top. This is the cleanest single piece of evidence in the index for the convention that vintage rather than marketing volume decides this axis.

The capability is being built out fast and the governance surface has not followed. The AI Assistant launched March 2026 executes outbound workflows from natural-language instruction, and Apollo's own engineering job descriptions name Autonomous AI Agents, Deep Research Agents and multi-agent systems that reason, plan and execute complex sales workflows.

Against that: no product documentation on approval gates, human review before send, escalation thresholds, guardrails, or an audit trail of AI-initiated actions. Worth distinguishing carefully, because it is easy to miscredit: Apollo publishes a strong editorial piece on human-in-the-loop AI selling that names specific escalation triggers (high-value accounts, negative sentiment, competitor mentions, contract discussions). That is thought leadership addressed to the industry, not a description of controls shipped in the product, and it is graded as such.

No model provider, family or version named for the AI Assistant, the email writer, the research agents or the semantic search layer. No model cards, no evaluation results, no accuracy or hallucination measurement published for any AI output. The AI surfaces are described entirely by what they do.

Notable as a gap rather than an omission of scale: third-party compliance vetting specifically records that Apollo does not hold ISO 42001, the AI management system standard, which is currently the closest thing to a certifiable answer on this axis and which no vendor in this index yet holds.

Heavily used and independently rated, with the headline claims contested. Apollo publishes scale figures (550k+ companies, 96% email accuracy) and third-party ratings exist at volume (G2 scoring data accuracy 8.3 of 10, Capterra scoring support 4.2 of 5). Two debits.

First, the contact count is inconsistent across Apollo's own surfaces - 210m in its careers copy, 224m in its insights content, and 230m and 275m in figures quoted from its marketing elsewhere - which for a company whose product is a count of records is an odd thing to leave unreconciled. Second, independent reviews across G2, Reddit and Trustpilot consistently report bounce rates of 15 to 25 percent, which is hard to square with a published 96% email accuracy figure.

The credit that keeps this at B and is genuinely rare: Apollo's own knowledge base publishes faq entries titled "Why do third-party email verification vendors flag Apollo's verified emails as invalid?" and "Why do I have a high bounce rate for emails sent via a mailbox connected to Apollo?" A vendor documenting its own most common quality complaint, in its own help centre, is disclosure of a kind almost nobody in this index offers.

The deepest regulatory registration surface in the index, and it is registration rather than assertion. Apollo is a registered data broker in California, Oregon, Texas and Vermont and names all four; it publishes its legal entity and address; it has an appointed Data Protection Officer with a published contact address; it names an EU Representative (Lionheart Squared) as Article 27 requires; and it is registered under the EU-US Data Privacy Framework including the UK Extension and the Swiss-US framework.

The shipped control that makes this an A rather than a paperwork exercise: Apollo screens its database against multiple do not call registries in-product, so the compliance obligation is discharged at the data layer rather than left to the buyer to remember. Add a published route for any individual to claim or remove their own profile. Naming the statute, shipping the control and publishing the removal path is the Aloware A pattern, applied at scale.

The most complete privacy disclosure in the GTM index. Apollo states plainly that it is GDPR-compliant as both controller and processor and, unusually, explains why - it is a controller of the data in its contributory database while that data sits in the database, and a processor of the customer data it handles in delivering the service.

A Data Processing Addendum is published at a stable URL and incorporated by reference into the Terms of Service, covering both roles and carrying the EU Standard Contractual Clauses for transfers. Data Privacy Framework certification covers EU-US, UK Extension and Swiss-US. A Privacy Center gives data subjects a self-service route to exercise rights.

The processing notice enumerates the categories held - role, employer, title, professional responsibilities, office location, professional social profile, job history - and discloses that Apollo makes inferences about those contacts, including their likelihood of interest in a customer's product. Publishing the existence of inferred data, which is exactly the category most brokers leave unmentioned, is the single strongest disclosure decision on this record.

First A on this axis in the index, and it resets the bar for every data vendor that follows. Apollo names three sourcing routes and does not hide behind "multiple sources": (1) the data contributor network of over two million contributors who share their business contacts in the course of using Apollo, with the specific extracting features named - Email Finder and Open Tracker - so a buyer can see exactly which product surface turns their mailbox into inventory; (2) publicly available sources; (3) purchased lists.

It accepts controller status over that contributory database. It is a registered data broker in four states. And it discharges the article 14 obligation in a published artefact - a GDPR processing notice whose opening line is that this is Apollo letting you know it has added your business contact information to its database, followed by the legal basis, the categories held and the rights available.

Compare Adapt.io, held at B because contribution carried no documented opt-out; Apollo's contributor network is described as opt-in and profile removal is self-service. The hard edge that still belongs in any comparison: the individuals whose details are extracted from a customer's connected mailbox never agreed to anything - the customer did. Apollo's answer is the notice and the removal route, which is the best answer available, and it is still an answer to a problem the business model creates.

Split by platform, and the split is instructive. On Google, Apollo holds CASA tier 2 certification (Cloud Application Security Assessment), the formal third-party conformance regime for applications requesting restricted Gmail scopes - a stated, independently assessed conformance position with a named platform, which is precisely what this axis reserves its top band for. Sending runs on Apollo's own infrastructure through official Gmail and Outlook OAuth connections, not scraped sessions.

On LinkedIn, nothing: the Chrome extension pulls contact data from professional network profiles, third parties characterise the sourcing as including web crawling, and Apollo takes no published position on LinkedIn's User Agreement. An A here would require the Google-side rigour extended to the platform Apollo actually has exposure on. B is the honest average of a certified position on one platform and silence on the other.

The training boundary is unaddressed, and at Apollo the question compounds in a way it does not elsewhere. Nothing published states whether customer CRM records, sequence content, call recordings or connected-mailbox content are used to train or improve models serving other accounts. The compounding: Apollo already discloses that connecting a mailbox feeds business contact data into a shared contributory database served to every other customer, including competitors.

A buyer has therefore already accepted one channel by which their data flows outward, which makes the unanswered question of whether a second channel exists for model training materially more pressing than at a vendor with no contributory model at all. The disclosure that exists covers the data network; it does not extend to the AI layer built on top of it. What counts in favour: encryption at rest and in transit, multi-tenant isolation enforced at the application and API layer, a uniform authentication policy, and a dedicated security team.

A useful distinction this record makes clearer than any other in the index: Apollo discloses that it holds you; it does not disclose that the message you received was machine-written. On the first, it leads the field - a published Article 14 processing notice sent to the standard of telling a person their data was added, a Privacy Center for exercising rights, and a self-service route to claim or remove a profile from the database.

On the second, silence: the AI Assistant and AI email writer generate outreach delivered to individuals, and there is no Article 50 position anywhere, no statement that AI-generated messages carry any marking, and no described mechanism by which a recipient could know. transparency and AI-nature transparency are different obligations under different articles, and Apollo has discharged one and not the other.

Broad and practical rather than deep and documented. Bi-directional CRM integration with Salesforce and HubSpot, Gmail and Microsoft Outlook via OAuth, a widely used Chrome extension, Zapier, and a large connector catalogue. API access exists but is gated to the organization tier and above, which puts programmatic access out of reach for the self-serve buyers the free and Basic tiers are designed to attract.

The March 2026 Pocus acquisition folds signal-based revenue intelligence into the platform, though packaging within the existing tiers has not been announced. Held at B: no object-level field mapping published, no MCP server or agent-gateway endpoint located, and the depth of any individual connector is not documented.

A single-region posture with no options in it, disclosed in the DPA rather than in marketing: Apollo hosts the service with outsourced, US-based data center providers on multi-tenant infrastructure, with the physical and environmental controls of those providers audited under SOC 2 Type II, and contractual agreements and vendor compliance programmes stated as the assurance mechanism for sub-processors.

Same band and same reasoning as 9Lenses and Amplemarket: clear disclosure of a posture with no customer choice beats silence. Held at B rather than higher because the disclosure is thin on specifics - no cloud provider named, no region named, and no EU residency option offered despite Data Privacy Framework certification and a substantial European customer base.

Real certifications, a real trust centre, and the weakest gating pattern on the report itself. Present: SOC 2 Type II and ISO 27001, both with the auditor named (A-LIGN), which is more than most vendors disclose; CASA Tier 2; a trust centre at trust.apollo.io carrying the compliance set (SOC 2, ISO 27001, GDPR, CCPA, CPRA, EU-US DPF, CASA Tier 2) with downloadable documents including a penetration test report; encryption at rest and in transit; multi-tenant storage isolated at the application and API layer; a uniform password policy; a dedicated security team.

The reason this is not an A: the security page's instruction for the complete SOC 2 report is to contact the sales team. The index bar is that a self-service report earns A and a gated one earns B, and routing the core security artefact through a sales conversation rather than a trust-centre request form is the most friction any certified vendor in this index puts on it.

The benchmark for this axis in the index. Five tiers published with real numbers and real units: Free at $0 with a genuinely usable allocation (full contact database access, roughly 900 credits a year, 2 active sequences, the Chrome extension) rather than a crippled trial; Basic $49 per user per month annually or $59 monthly; Professional $79 or $99; Organization $119 or $149 with a three-seat minimum; Enterprise custom. The annual discount is quantified at roughly 17 to 20 percent.

Credit allocations, sequence limits and feature gates are published per tier, so the buyer can see that AI features are absent from Basic and that SSO and API access begin at Organization. A buyer can budget their entire self-serve purchase path and knows precisely what the enterprise conversation is for, which is the established test.

One item that is not published and belongs in the note: third-party procurement sources report contracts commonly carrying 5 to 10 percent annual price escalation clauses. The contrast worth keeping: Apollo publishes five tiers; Adobe and Salesforce publish none.

The finding here is structural and worth reusing across every credit-metered data vendor: export is a purchased commodity, not a right. Export credits are allocated per tier as a commercial unit - ten a month on the free plan - so a customer's ability to leave with the records they have already paid to reveal is rate-limited by the plan they are on, and a large book of enriched contacts cannot be extracted faster than the subscription permits.

CSV export and API access (Organization and above) are the mechanisms. Nothing published covers post-termination customer data rights, a deletion timeline or confirmation, what becomes of sequence and engagement history, or what happens to data derived from a connected mailbox after churn. Credit for the side, which is genuinely good: any individual can claim or remove their own profile from the database self-service. That is portability for the person in the database, not for the customer who bought the seat.

First D on this axis and the sharpest contrast in the index: Apollo and Amplemarket sell into the same buyer and land at opposite ends of this axis in the same session. No warmup, no inbox placement testing, no domain health monitoring, no mailbox rotation policy, no published bounce or complaint threshold, no volume governance and no documented response to reputation degradation - on a platform built for high-volume sequencing from the customer's own connected Gmail or Outlook mailbox, which means it is the buyer's domain reputation being spent.

The load-bearing evidence is Apollo's own: its knowledge base publishes faq entries addressing why third-party verification vendors flag Apollo's verified emails as invalid, and why customers see high bounce rates from mailboxes connected to Apollo. Independent review consensus across G2, Reddit and Trustpilot puts bounce rates at 15 to 25 percent.

A competitor's published analysis states Apollo discontinued email warmup in 2024, leaving no native deliverability protection; that claim is from a rival and is treated as third-party rather than established, but nothing on Apollo's own surface contradicts it or documents any replacement. Honesty about the failure mode is real and is credited on the outcome evidence axis; it does not substitute for the discipline itself.

The tier ladder maps the segment coverage explicitly and it runs the full width: a free plan aimed at individual users and light prospecting, Basic and Professional for small teams, Organization for larger sales orgs with a three-seat floor, and Enterprise for buyers needing SSO and dedicated support. Apollo claims 550K+ companies using the platform, and the database is described as worldwide across contacts and companies.

Held at B: no enumerated country coverage or per-region data density is published, and independent reviews consistently report that data accuracy varies materially by region - which is precisely the coverage detail this axis asks for and the vendor does not supply for a database sold on its global reach.