Firstlines.ai

The price list is the evidence and it is unusually stark here. Neither priced tier carries a single generated or model driven feature: the two plans differ from each other only by campaign count, and every line item on both is mechanical, being connected account, safety limits, notifications, list import and support. A banner announces sequencing powered by models as new, and the vendor's own domain name is itself the claim.

Remove the model entirely and what remains is the complete product the tiers actually sell, which is a connection request and follow up engine. The reference test in this index is to go to the price list before the marketing, and the price list has no model in it.

The product is sold as running on auto pilot and the privacy notice states plainly that the service executes automation sequences, meaning connection requests and messages, on the operator's behalf. Multi step campaigns with delays and follow ups run without a person in the loop once configured. No approval gate, review step, audit trail, withholding guardrail or escalation path appears anywhere.

The only constraint published is a daily volume limit, which governs how much the system does rather than whether any individual action should happen, and the terms describe that limit as a measure to mimic human behaviour rather than as restraint on the agent.

A banner announces sequencing powered by models as a new capability and the vendor's domain carries the claim in its name. Beyond that there is nothing: no model, provider, family or version named, no description of what the model does in a sequence, no accuracy or quality measure, and no fallback behaviour. Neither the privacy notice nor the terms mentions model processing at all, so a buyer cannot tell whether prospect data reaches a model, which model, or under what terms.

A trust bar claims more than a thousand sales teams and displays five customer names. Two of them are recognisable products in this market. The other three read as the generic placeholder names that ship inside website templates, and none of the five carries a logo image, a link, a customer page or any other corroboration. No named individual, no title, no case study, no quantified outcome and no independent review platform record was located.

A demonstration video is embedded and is the only evidence of the product working. The gap is not a production problem: the site is current, the legal documents are dated and maintained, and the pricing is complete, so the absence of any customer evidence sits alongside a surface that is otherwise looked after.

The finding here is an internal contradiction rather than an absence. The acceptable use section of the terms prohibits the customer from sending spam, unsolicited messages or harassment through the service, on a product whose entire purpose is sending unsolicited connection requests and messages to strangers at scale.

The same section prohibits importing data the customer has no right to use, which places the lawful basis question on the buyer without giving them any means of answering it. Beyond those two prohibitions no consent standard, opt out mechanic, suppression capability or regulation of any kind is named. The vendor also disclaims responsibility for the customer's compliance with the platform's own community policies, so both the platform duty and the data duty are assigned to the buyer.

Current at October 2025 and correct on the one allocation this category usually gets wrong: prospect data is expressly assigned to the customer as controller with the vendor as processor. Infrastructure is named, being a hosted database platform and a payment provider. Adherence to the extension store's limited use requirements is stated, the vendor states it does not sell data, and a non affiliation statement appears.

What holds it at this grade is the list of things a reviewer asks for that are simply absent: no legal basis for any processing, no retention period at all, no international transfer mechanism, no supervisory authority, no officer, no portability right and no sub processor list beyond the two named platforms. The governing law is stated as Hong Kong while no company entity, registration number or address appears anywhere on the site. One clause is template residue and worth recording, being an assurance that data is not used for lending or credit scoring, which belongs to a financial services document rather than this one.

At the bottom of the band, and what earns it is an allocation most of this category never makes: the vendor states in writing that the customer is the controller of imported prospect data and that the vendor is only the processor. No standing inventory of records is held or resold, no purchased database is offered, and the raw material is what the operator imports from their own network or a file.

Held at the bottom because the allocation resolves the vendor's position without helping the person in the data: no notice, lookup or removal route exists for a prospect whose name, title and employer sit in the system, and the terms push the entire right to use question onto the buyer while providing no means to establish it.

At the top of the band, and the vendor supplies the evidence in its own terms, which contain the most explicit risk allocation of any vendor graded on this axis. It states that the customer is solely responsible for complying with the platform's user agreement and community policies, that the use of automation tools may be against that platform's terms of service, that it cannot guarantee the account will not be restricted, flagged or banned, and that the customer uses the service entirely at their own risk.

Account restriction is then named specifically inside the liability exclusion list. The architecture underneath is what places it here. The extension takes the operator's session cookies so the platform can act as them, execution is autonomous rather than approved action by action, and the vendor's own description of its safety measures is daily limits and random delays to mimic human behaviour, which is evasion stated as mechanism.

Two facts keep it at the top rather than lower and both matter: no account rotation or multiplication is sold, with one connected account on each priced tier and credential sharing prohibited, and nothing celebrates undetectability as a benefit. Naming the exposure this fully is better than concealing it and it does not cure it.

The training question is unaddressed in every direction. No statement covers whether operator content, imported prospect records or message history feed any model, no tenant boundary is described, and no model provider is named to which such a question could attach. The nearest clause is a general permission to use collected information to improve, personalise and expand the service, which is the shape that leaves the question open rather than answering it. The service holds session credentials and message content for its users, which is what makes the silence consequential.

At the bottom of the band. Nothing adopts a false identity: the connection request and the messages come from the operator's own real account under their own name. What the recipient cannot see is that no person chose the moment or, in a sequence, the words, and that the timing was randomised specifically so the pattern would not read as machine generated.

The vendor states that mechanism itself, describing random delays intended to mimic human behaviour, which is manufactured effort described by the party manufacturing it. Automated profile visits carry the same problem in a milder form, since the whole communicative value of a profile view is that somebody chose to look. Article 50 of the European artificial intelligence regulation goes unmentioned. Held at this grade rather than lower because the mimicry appears in a risk disclosure rather than as a marketing boast, and because no undetectability claim is sold as a feature.

What exists is a browser extension published on the official store and file based list import on both tiers. What does not exist is everything the site advertises: the footer lists documentation, a help centre, an application programming interface, a security page, a features page, an about page and careers, and every one of them is plain text rather than a link.

No connector to any system of record is named anywhere, so the meetings and replies this product generates have no described route into a pipeline, and the customer's own records cannot flow in except through a file.

Two infrastructure providers are named, a hosted database platform and a payment processor, which is more than silence and is the reason this is not lower. No country, region, provider region or residency option is stated for any of it, and no sub processor list exists beyond those two names.

The jurisdictional picture is unusually opaque even by the standards of this category: the terms nominate Hong Kong law while no company entity, address or registration number appears anywhere on the site, so a buyer cannot identify where the counterparty sits or where the session credentials it holds are stored.

The service receives and stores the operator's professional network session cookies so it can act as them, which is the most sensitive access any product in this category requests. Against that, no certification of any kind is claimed, no auditor, report, audit period, trust page, status page, penetration test or enumerated control set was located, and the footer advertises a security page that is not a link.

The whole published security position is one sentence naming transport encryption and two hosting providers, followed by a statement that no security system is impenetrable. Graded on the precedent already set in this category for a vendor holding live authenticated sessions with no documented control set, and applied here rather than reasoned fresh.

Two tiers carry real numbers with a real unit, at 39 and 79 dollars a month with a monthly and yearly toggle stating four months free on annual billing, a seven day trial requiring no card, self serve registration and cancellation available in account settings.

The terms add the mechanics most vendors leave out: billing in advance, cancellation effective at the end of the current cycle, and an explicit statement that fees are non refundable with no credit for partial months or for downgrades. A buyer can budget the purchase without speaking to anyone. Two things hold it off the top.

The meter that actually governs this product is daily action volume, and it appears on both tiers as an unquantified phrase about smart safety limits with no number attached anywhere. And the two priced tiers publish near identical feature lists differing only in campaign count, so the page does not explain what the extra forty dollars buys beyond running more than one campaign at a time.

Governing terms are published and current, cancellation is self serve and takes effect at the end of the billing cycle, and the privacy notice grants a deletion right on request. That combination keeps this above the bottom of the scale. What is missing is the working half. No retention period, post termination window, deletion timeline or deletion artefact appears in either document.

And the asymmetry is concrete rather than theoretical: file based list import is a named feature on both priced tiers, and no export counterpart is named anywhere, so the campaigns, sequences, message history and accumulated prospect records have no described way out.

No electronic mail is sent, so most of this axis is inapplicable and the grade records what is knowable with that context. One pacing control is published and it is named rather than specified: smart safety limits appear on both tiers with no daily figure, no ramp, no per account rate and no stated behaviour when a limit is reached.

The terms describe the same mechanism as daily limits and random delays whose stated purpose is to mimic human behaviour, so the only pacing discipline in the product is aimed at the platform's detection rather than at the recipient's experience or the account's long term health.

At the top of the band. Each tier card names who it is for, being individuals getting started, professionals scaling outreach, and teams managing multiple accounts, and the top tier is built around an agency motion with white labelling, a multi account dashboard and a named success manager. That is a real if shallow segmentation and it is consistent across the page.

What is absent is everything that would let a buyer place themselves precisely: no headcount band, no role definition beyond the tier captions, no vertical, no geography, no stated ceiling, and no company entity or address published anywhere, so even the vendor's own market is unstated.