Addy AI Review: promising focus, serious gaps (52/100)

Fact	Value
Domain	addy-ai.com
Category	AI workflow automation for mortgage lending and loan origination
Pricing	Unknown; price range not supplied
Pages crawled	15 (13 dimensions in scope; 2 excluded)
Crawl date	2026-10-10

Addy AI has a clear market position: it targets mortgage professionals with a concrete speed-and-automation promise, and named customer roles plus quantified testimonials provide useful proof. Audience and messaging score 74/100, while first impressions and positioning score 78/100.

The review weakens where execution must carry that promise. Usability scores 58/100 and accessibility 58/100. Performance scores 42/100 after a 36.9 MB mobile lab transfer and slow lab rendering. Writing quality scores 48/100. Risk and stability score 38/100, editorial QA 42/100, technical SEO 38/100, and Docs and self-serve help 20/100.

Addy AI is presented as AI automation for mortgage lending. The hero promise, “Close mortgage loans in hours, not weeks,” is specific, and named customer roles with quantified testimonials add useful proof. The 90%, 99.99%, tenfold, and 4x claims need definitions and methodology. The supplied messaging addresses loan officers, branch managers, and mortgage companies, so the primary buyer is not narrowly defined. Choose one canonical brand domain, align internal links, and add evidence notes to performance claims.

Addy AI gives mortgage visitors a recognizable path from workflow promise to features, proof, and a demo. SOC 2 Type 2 messaging, customer logos, and named testimonials support trust. The message addresses loan officers, branch managers, and mortgage companies together, so the primary buyer remains unclear. The case-studies page further weakens information scent with stale AI and cloud metadata. Add persona-specific entry points and link each quantitative claim to customer context or methodology.

The homepage communicates the mortgage-AI offer, but key follow-up tasks are not reliably complete in the captured artifacts. /contact-us has no captured form, button, phone, or email target, while /blog is a graph dead end with no substantive captured article content. Mobile LCP is 7.6 seconds and TTI is 13.9 seconds. Replace generic Learn More labels with product-specific labels, make contact action explicit, and expose substantive blog content.

The available accessibility evidence scores 58/100 and shows material barriers. Lighthouse repeatedly failed aria-hidden-focus, meaning focusable content is hidden from assistive technology; the homepage also lacks a main landmark, while link-name and color-contrast checks fail. CSS evidence contains one outline:none and no focus or focus-visible rules, so keyboard visibility needs verification. Remove aria-hidden from interactive containers, add semantic structure and names, restore visible focus styling, and retest contrast and keyboard order.

The supplied screenshots show a designed marketing surface, but they do not provide enough measurement to verify contrast ratios, tap-target dimensions, heading structure, scroll width, input sizing, or CLS. The 68/100 score is therefore provisional. Run computed DOM and CSS checks on the homepage, contact, and blog templates, then validate representative product, pricing, and form templates at 390px before making site-wide claims.

Performance is mixed: field data is fast, but the mobile lab exposes a severe payload problem. CrUX reports CLS of 0, while the mobile lab reports 36,980 KiB transferred across 115 requests. Framer assets contribute 36,873 KiB across 79 requests, and the page has 29 blocking requests.

The homepage has concrete promises, including “Close loans 90% faster” and “Submit CTC-ready loans in 5 minutes,” but the sampled page is reported as one 610-word paragraph with 29-word average sentences. The blog and 13 sampled routes expose generic text, including “Addy AI - Stealth Startup in San Francisco,” with no H1. Split the homepage into scannable sections and give every live route a unique H1, title, description, and value proposition.

The evidence supports a technical exposure diagnosis, not a measured traffic explanation. The addy-ai.com canonical points to addy.com through a two-hop redirect, no sitemap was found, and a random probe returns a 200 soft 404. Only 15 of 40 pages were crawled and Search Console data is not supplied, so traffic loss and rankings are not measurable here. Repair canonical routing, indexability, sitemap coverage, and route output before interpreting demand changes.

Editorial readiness is held back by repeated route metadata and dense homepage structure. The crawler reports the same title on 13 pages, a shared meta description on 13 pages, and no H1 on generic-shell routes; the homepage is one 610-word paragraph with 29-word average sentences. Complete or redirect generic routes, give each indexable page distinct metadata and an H1, add provenance beside quantified claims, and run a structural edit on the homepage.

The captured crawl does not credit a usable public documentation system. The /llms.txt route returns the same HTML shell rather than curated documentation. Publish a searchable help center with getting started, mortgage-workflow task guides, reference material, and a clear support path.

Technical SEO has a serious consolidation problem. The homepage canonical is addy.com, while addy-ai.com requests redirect through addy.so before reaching addy.com; no usable sitemap was found, and a random path returns HTTP 200 with generic content. Thirteen routes lack H1s, many expose only 8 words, and titles and descriptions are duplicated across 13 pages. Choose one host, redirect directly, publish its sitemap, return true 404 or 410 responses, and restore unique crawlable route content.

Addy AI is best suited to mortgage professionals evaluating workflow automation through a demo-led buying process. Its strongest assets are specific audience targeting, a concrete automation promise, named customer roles, and quantified testimonials; first impressions and positioning reach 78/100.

Resolve the canonical and redirect mismatch, absent sitemap, and soft-404 behavior behind the 38/100 technical SEO score. Reduce the 36.9 MB mobile lab payload and address accessibility failures. Then give buyers a usable public help system: Docs and self-serve help scores 20/100.

This 13-dimension review uses a crawl of 15 pages completed on 2026-10-10 plus the supplied public dimension score table and one-line summaries. The article frame does not re-score dimensions or add findings beyond those inputs.

Dimensions 12 (Decision-support surfaces) and 13 (Review-content integrity) were excluded as not applicable. Pricing is unknown in the supplied facts. The public inputs do not provide Google Search Console enrichment. Read How SiteList scores for the review method and dimension definitions.