What's Trending

Peel the model layer off and a working listening tool remains. Collection across public sources, retained post links with author and timestamp, an engagement dashboard, and a ranking score the vendor states is computed from activity, unique authors, recency, platform breadth and available public engagement, which is arithmetic over counts rather than inference.

The model driven parts are narrower and the vendor does not overstate them: suggested market terms and competitor profiles derived from a submitted website, topic summaries, and the content directions it calls planning aids. Its own questions page states plainly that the product does not publish content for the customer and that the buyer decides what to create and where. The claim is proportioned to what was built, which is why this sits at the conventional platform band rather than below it.

The outward autonomy boundary is that there is none: the product observes and ranks, and nothing leaves the workspace under the customer's name. What runs unattended is scanning, on schedules as frequent as hourly on the top tier, plus scans an external agent triggers through the published tool endpoint.

Containment on that surface is documented rather than asserted: keys are scoped to a single workspace, displayed once, stored as hashes and revocable in settings; an agent triggered scan is rate limited to one per hour per workspace; and the vendor enumerates what a key cannot reach, naming other workspaces, private platform data, billing operations and stored credentials.

Guidance published for agent authors tells them to treat generated content ideas as briefs requiring human review and to reuse stored evidence rather than trigger another scan. Held below the top band because the vendor states that an agent triggered scan mutates workspace history while publishing no audit trail, no record of which key acted, and no notification to the account owner when it does.

Ranking is explained and the model layer is not. The vendor publishes the five inputs behind a topic score, states that a score is a prioritisation aid rather than proof of total market demand, tells readers that a missing metric means unknown rather than zero, and instructs agents to separate observed metrics from inferred significance. That is more explainability of scoring than most vendors in this index publish. Against it, no model, provider, family or version appears anywhere.

The privacy policy refers to artificial intelligence processing providers as an unnamed category, the workspace holds a model connection setting that is never described, and the topic summaries and content directions are generated text whose provenance is unstated. Documenting the generative half at the level the deterministic half already reaches would move this.

Customer evidence is absent in every form: no named customer, case study, testimonial, logo wall, review count or usage figure anywhere on the site or in its sitemap. What holds this at the middle band rather than lower is that nothing was fabricated to fill the gap. The homepage workspace panel carrying post counts, author counts and engagement figures is labelled an illustrative workspace, which is the honest form of a pattern this index has repeatedly recorded in its dishonest form.

The vendor's answer to proof is a complete first scan run free with no payment card, which puts the burden of demonstration on the product rather than on a claimed statistic. One named customer with a stated period and metric would move this.

This axis barely bites here, because the product sends nothing: there is no email, dialer, message or connection request in it, only collection of public posts, ranking and source links returned to the buyer. What is knowable is that the terms of service place lawful use and respect for third party rights and platform rules on the customer, and that nothing in the product is described as enforcing any of it. Recorded at the responsibility transferred band with the structural context carried in this note, per the convention for products in this index that do not send.

Two halves, and only the customer facing one is answered. For the account holder the policy is current, dated 20 August 2026, and specific: analytics providers named individually, with Mixpanel on its European data residency service and address enrichment disabled, Microsoft Clarity and Sentry named for interaction and error diagnostics, an accept or decline choice for optional analytics, a statement that passwords, credentials, keys, provider tokens and raw sensitive form values are never sent to analytics, compliance with the Google limited use requirements stated by name, no sale of personal information, and a route to request access, correction, export or deletion.

For the people whose public posts become the evidence, which is the personal data this product processes at scale, there is no lawful basis, no notice, no objection route and no retention period. There is also no processing agreement anywhere, no sub processor list, no named legal entity in the policy and no supervisory authority route, and neither European nor California privacy law is named on any page read.

Provenance is described platform by platform, which is rare at this size and rarer at this price. The published reference states that Reddit is read through public search feeds with a public archive fallback, Facebook through configured public page feeds where accessible, Instagram through Meta's authorised professional account and business discovery surfaces, and X through configured official access when present with public profile and embed fallbacks otherwise.

The vendor states outright that it never claims access to private posts, private comments, private impressions or metrics a source does not expose, and the collected fields are enumerated: platform, author, text, original link, publication time, matched terms and source specific public metrics. Held below the top band because the legal footing is never stated. There is no licence position, no terms of service position, and no indemnification, and the unnamed public archive standing behind the Reddit fallback is precisely where that question bites.

The operating method is disclosed rather than implied, and most of it is sanctioned: Meta's authorised professional account and business discovery surfaces are named for Instagram, configured official access is named for X, and configured public page feeds for Facebook. Nothing in this product rents or rotates accounts, ships a browser extension, automates a logged in session or markets evasion, and the terms instruct customers to respect platform rules.

The unresolved ambiguity, and the reason this is not the top band, is that the same reference describes fallback paths for when official access is absent, naming public profile and embed fallbacks for X and a public archive for Reddit, and neither fallback carries a conformance position. The vendor also states that coverage and available engagement fields vary with platform availability, which is a continuity risk the buyer inherits rather than one the vendor controls.

The training question is answered for one narrow input and left open for everything else. Google account information is stated not to be sold, not used for advertising and not used to train generalised models, which is precise and precisely limited.

Nothing states whether workspace configuration, collected evidence, generated reports or submitted market terms reach the unnamed artificial intelligence processing providers, whether those providers retain them, or whether anything improves a model serving other workspaces.

Retention is described by purpose rather than by period, with account and workspace data kept while the account is active and as needed for security, legal and operational purposes, and cached public posts refreshed or deleted as sources change. Deletion is available by written request with no stated timeline.

This axis has no outbound surface to grade, since the product contacts nobody: it does not send, call, message or converse, and it carries no persona or sender identity. The one live question it raises is the inverse of the usual one. Identifiable individuals, their handles, their post text and their published times are collected, stored and ranked without notice to them, and the resulting content direction is aimed back at the audience those posts came from.

Nothing published addresses whether an individual can object to their post being retained as evidence, and the disclosure obligations in force since August 2026 are not mentioned anywhere. Credited in the note rather than the grade: the product preserves author attribution and the original source link on every item and instructs agents to show source links beside conclusions, which is the grounding discipline this axis rewards in products that do generate outreach.

The programmatic half is unusually complete and the connector half does not exist. Published and readable without an account: an interface contract in OpenAPI 3.1, four documented read and scan endpoints, a stateless tool endpoint exposing three named tools with discovery metadata at a well known path, an agent manifest, an agent skill file, separate markdown guides for the interface, authentication and pricing, and a consolidated machine readable product reference.

Access is priced into the fifty dollar tier rather than reserved for an enterprise conversation, which is the opposite of the pattern this index records at larger vendors. Against that, there is no connector to anything a marketing team already runs: no system of record, no marketing automation platform, no workflow tool, no calendar, no chat destination, no webhooks and no marketplace listing. Evidence leaves this product only by being read.

Hosting is answerable only as cloud. The marketing site, the application and the interface run as separate hosts on one domain, and nothing states the infrastructure provider, region, country or tenancy model for the store where workspace configuration and collected evidence live. The only residency statement anywhere concerns product analytics rather than customer data, naming Mixpanel's European data residency service. For a buyer base that includes agencies handling European client material, the residency question is left to a conversation that has no published starting point.

A small but real control set, and no independent verification of any kind. What is published: interface keys stored as hashes, displayed once, scoped to one workspace and revocable in settings; a requirement that a key stays server side and never appears in a query string or on a public page; a statement that credentials and keys are never sent to product analytics; named error monitoring; and a crawler policy that disallows the interface path.

Beyond that there is no security page anywhere in the sitemap, no certification, no audit or report, no penetration test, no trust centre, no status page and no vulnerability disclosure route. The sensitivity class sits below the sending tools this index grades, since the product holds no mailbox credential and no contact database, though it does hold sign in identity and, where the keyword metrics feature is enabled, delegated advertising interface access. One completed audit with its type and period stated would move this.

A buyer can compute the entire bill from published pages, and the pricing reference is itself machine readable and dated. Three tiers with no quote only step: a free complete first scan with no payment card, $20 per month or $192 per year for one full scan every seven days with a seven day trial, and $50 per month or $480 per year for unlimited scans with hourly scheduling and agent access.

The annual equivalent is stated for each tier, the limits are enumerated tier by tier rather than implied, the rate limit of one agent triggered scan per hour per workspace is printed where a buyer sees it, and the billing mechanics are published in full: monthly and annual renewal until cancelled, immediate upgrade with provider calculated proration, downgrade scheduled to the next billing date, checkout total displayed before payment, and cancellation effective at the date the billing portal shows. Two gaps worth naming, neither load bearing: refund terms are unstated, and no scan is quantified by the volume of posts or sources it collects, so the unit of value behind the price is uncounted.

Renewal and cancellation are documented better than most of this index, and offboarding is not documented at all. Published: monthly and annual subscriptions renew until cancelled, cancellation stops future renewal at the effective date the billing portal displays, a downgrade takes effect at the next billing date, and a written route exists to request access, correction, export or deletion of account data.

On the top tier the read endpoints are themselves a portability path, since stored evidence and ranked topics can be pulled in bulk through the interface. Missing is the part a buyer needs before signing: no self serve export or file format on any tier, no statement of what happens to stored evidence and scan history after cancellation, no deletion timeline attached to the request route, and no post termination data rights in the terms, which are silent on the subject and name no governing law.

This product has no sending surface, so the axis records what is knowable rather than a sending practice. Collection side discipline is the closest analogue and it is published: scans through the interface are rate limited to one per hour per workspace, the middle tier is capped at one scan every seven days, agent guidance instructs reuse of stored evidence before triggering a fresh scan, and the vendor states that coverage and available engagement fields vary by platform rather than promising completeness. Unpublished on the same subject: request rates against the platforms themselves, and what the product does when a source rate limits, blocks or withdraws access.

Who it is for is stated clearly and evidenced by nobody. The buyer is named repeatedly as marketers, founders, lean content teams and agencies, and the site supports that with five use case pages and a set of agency oriented guides covering brand monitoring, competitor monitoring, content research and client reporting.

Against it: no customer is named anywhere, no customer count, no team size band, no region, and no language coverage statement at all, which is load bearing for a product whose function is reading conversations across five platforms. The company base recorded on this profile is London, stated by the vendor; the about page separately gives an Acton, Massachusetts address for Wavicle, the studio that operates the product.