Signal87 Review: focused product with a mobile gap (72/100)

Field	Value
Domain	signal87.ai
Category	Document intelligence and legal AI research SaaS
Pricing	Paid; USD 20–50
Pages crawled	14
Crawl date	2026-10-09

Signal87 earns 72/100 by presenting a focused document-intelligence product for legal, compliance, real-estate, and related governance teams. Its strongest evidence is concrete: citation-backed research, exact spreadsheet analysis, document examples, and a clear upload-to-answer workflow. The experience also has a strong semantic baseline, with an accessibility score of 86 in one PSI run and no horizontal scroll in the sampled mobile render. The central weakness is performance on mobile: the homepage took 7.3 seconds to show its largest content, compared with 1.3 seconds on desktop. Writing quality is fair at 68/100 because the product examples are specific but the presentation can be tighter. Risk and stability are also fair at 67/100; without a Search Console connection, organic traffic loss and query trends cannot be measured.

Signal87 makes its audience and core promise legible within five seconds: "For legal, compliance & real estate teams" and "Ask every document at once." The claim is concrete because the homepage shows source passages, a saved report, and the calculation "14 of 46 units expire within 12 months." That evidence gives the product a clear shape beyond a generic AI assistant. The proof is still mostly illustrative: the supplied material includes team names and program recognition, but no named customer outcome. Competitive framing is also implicit; no comparison pages were discovered. The strongest next move is to choose one primary buyer and attach a measurable customer result to that document decision.

Signal87 gives first-time visitors a coherent model: upload PDFs, Word files, and Excel; ask a plain-language question; receive an answer tied to its source files. The ARIA example and spreadsheet table make the workflow tangible, and the five-day free trial offers a clear next step. For legal and governance work, trust belongs beside the first upload action. The site provides privacy language, a Security link, citations beside report answers, and guidance to review important conclusions against originals, but the strongest institutional recognition appears later. Bring privacy, citation verification, and a concise explanation of exact computation versus generated narrative closer to the first CTA. The product is understandable; confidence at commitment is the remaining friction.

A first-time visitor can identify the product and audience quickly, then see concrete ARIA tasks, pricing, and "Start free" and "Contact us" actions. Trust destinations—Government pilots, Team, and Security—are reachable from the homepage, but a skeptical evaluator must visit separate pages to assemble the full proof story. The larger usability obstacle is mobile speed: the supplied lab run records LCP at 7.3 seconds, FCP at 6.2 seconds, Speed Index at 8.1 seconds, and Time to Interactive at 8.0 seconds. Prioritize the mobile hero rendering path, then add a compact trust summary beside the primary CTA. That should shorten the path to a first meaningful action.

Signal87 has a strong semantic baseline in the sampled pages: header, navigation, main, and footer landmarks are present; question-form controls have accessible names; and the supplied PSI run reports an accessibility score of 86/100. Three verified fixes remain. PSI failed the homepage color-contrast audit with a score of 0, so the affected foreground/background pairs should be measured against 4.5:1 for normal text and 3:1 for large text. The homepage heading sequence contains a second h1; demote it to the appropriate section level. No skip link was detected on the homepage, trial variant, terms, or privacy page. Add a first-focusable, visible Skip to main content link. Focus trapping and error announcements were not tested.

The visual system is structurally sound on mobile: the sampled page uses a sane viewport, 16px body and input text, and a 390px scroll width with no horizontal scroll. The principal design risk is delivery speed rather than layout. PSI records mobile FCP at 6.2 seconds, LCP at 7.3 seconds, and Time to Interactive at 8.0 seconds, versus desktop LCP at 1.3 seconds; it estimates 692 KiB of unused JavaScript. Fix the mobile critical path before polishing the interface. Desktop contrast also failed PSI. Finally, verify heading strings such as "Ask every documentat once." in the rendered DOM, because the crawl output shows collapsed spaces. Consolidating the reported 13 font-face declarations, 24 !important declarations, and 20 z-index declarations would keep the system easier to maintain.

Signal87 is fast enough on desktop in the supplied lab run, but mobile LCP is a clear performance risk: 7.3 seconds versus 1.3 seconds on desktop, against a 2.5-second target. The homepage transferred about 3.3 MB across 12 requests; PSI estimated 692 KiB of unused JavaScript, 530 ms of render-blocking savings, and 780 ms from multiple redirects. The exact LCP element was not supplied, so the fix should begin with a trace rather than an assumption about the hero asset. Identify and prioritize the mobile LCP element, defer non-critical work, inspect the five blocking requests, and consider serving right-sized image variants only if the trace identifies image sizing as a contributor. These are lab results, not field CrUX data; re-test toward LCP ≤2.5 s, INP ≤200 ms, and CLS ≤0.1.

Signal87's strongest editorial choice is demonstration: the homepage shows "The revised agreement extends the payment period from 30 days to 45 days" with source quotations, then makes the spreadsheet promise tangible with "14 of 46 units expire within 12 months." The weakness is presentation hygiene. The H1 is recorded as "Ask every documentat once.", other headings lose spaces, and the crawl reports 599 words in one paragraph. Verify the production DOM before treating the spacing as a rendering defect, then restore paragraph boundaries around the workflow, examples, use cases, pricing, and CTA. The crawler also found duplicate titles across four homepage variants and duplicate descriptions across four URLs. Canonicalize surviving variants and add a concise description to /terms.

A traffic diagnosis is not possible from the available first-party evidence: Google Search Console is not connected, and the Wayback enrichment returned no CDX rows. The technical risks that can be acted on are still clear. Probe results show multi-hop redirects from HTTP and apex HTTPS to the www host, while /government-pilots/ also uses a second redirect. Choose https://www.signal87.ai as the canonical host and make each alternate one hop. The sitemap lists 2 URLs while the crawl found 14 pages; compare that set with the intended public indexable pages before adding or excluding URLs. Mobile lab LCP is 7314 ms versus 1284 ms on desktop, but without field or Search Console data it is an experience risk, not proof of a traffic penalty.

Signal87 works best for legal, compliance, real-estate, legislative, and corporate-governance teams working across complex documents. The product earns trust by showing citation-backed research, a lease comparison, a sourced payment-term example, and an exact spreadsheet table. Its clearest fix is the mobile critical path: the 7.3-second largest-content result is well above the 2.5-second target, while desktop reaches 1.3 seconds. The next improvements are to make trust boundaries and verification clearer and tighten the presentation of already-concrete product copy. This is a capable, focused offer whose mobile experience needs attention before its strongest evidence can land consistently.

This 13-dimension review combines crawl evidence, public page content, simulated usability work, accessibility checks, and performance measurements available for the 2026-10-09 crawl. Signal87 was evaluated across the public score table; the overall article score is 72/100. Decision-support surfaces and review-content integrity were not applicable. Search Console was not connected, so organic traffic loss, query trends, and algorithm-update correlation remain unmeasured. See How SiteList scores for the review method.