0043 — SEO and AI-visibility strategy for the marketing site

• Status: Proposed
• Date: 2026-06-02
• Deciders: Fede (+ Sean for positioning/copy)
• Related: SEO roadmap (post-indexing): rank + appear in search (#1182)

Context

Until 2026-06-01 the entire public marketing site was noindex (a global X-Robots-Tag in next.config.js); Google had indexed nothing. Phase 0 (PR #1786) removed the block. As of 2026-06-02 the foundation is verified sound and the site is indexed:

Three forces shape what comes next:

1. Brand collision. "PropFlow"/"PropFlo" is shared by 7–8 other entities, several adjacent (observed via manual Google searches for "propflow" and "propflow ai", 2026-06-02: propflo.ai CRM, propflow.io, Atlas Energy's "PropFlow" proppant system, propflow.pro, propflow-ai.org, propflow.to, propflowtrading.com, propflowsolutions.com). A bare "propflowai" query auto-corrects to "propflow" and surfaces the namesakes. Branded-term disambiguation matters more for us than for a uniquely-named competitor.
2. Head terms are owned by incumbents. Category queries ("ai property management", "ai-driven solutions for multifamily operators") are dominated by RealPage, Yardi, EliseAI, Entrata, AppFolio — domains with years of authority and thousands of backlinks. A freshly-indexed, ~zero-authority domain cannot rank for these in days or weeks.
3. AI Overviews / LLM answers are "still SEO." Per Google's AI-features guidance (accessed 2026-06-02), there is no special schema, llms.txt, or markup that places a site in AI Overviews. Inclusion is downstream of being indexed + well-structured AND being cited by the third-party sources LLMs trust (G2, Capterra, "top X software" listicles, proptech press). We have none of those citations yet.

This ADR records how we sequence the work and — equally important — what we deliberately will not do.

Decision

Pursue search and AI visibility along three tracks, sequenced by ownership and risk, and treat brand-entity disambiguation as the through-line:

Track A — Engineering (autonomous PRs, low-risk, this week). One concern per PR.

1. JSON-LD structured data — Organization + SoftwareApplication + FAQPage. Rationale: not a path into AI Overviews, but how Google builds the brand entity / knowledge panel, which is our primary lever against brand collision. Additive; no copy changes.
2. Core Web Vitals — the homepage hero is a ~6MB autoloaded video that is (almost certainly) the LCP element. Fix: optimized poster image as first paint, preload="none", lazy-start on idle/in-view, re-encode the video to <1MB with modern codecs, reserve layout space to avoid CLS. Verify with Lighthouse/PSI on the preview before/after.
3. Measurement — GA4 + Bing Webmaster; GSC already verified with Request Indexing submitted. Query data populates ~1 week out.

Track B — Content & positioning (needs copy sign-off; weeks). 4. On-page keyword mapping of /platform, /solutions, /who-its-for; each section opens with a direct, factual answer (the format AI Overviews quote). 5. Comparison + "alternative-to" pages for the top-3 realistically-shortlisted competitors (EliseAI, Funnel, Colleen) — ~6 pages; these double as ranking pages and as the content LLMs cite for category answers. 6. Blog cadence: question-based, how-to, comprehensive topic overviews.

Track C — Off-site authority (mostly not code; ongoing, the real AI-Overview lever). 7. Directory listings — G2, Capterra, Product Hunt category pages. 8. Listicle placement — "best multifamily AI software" roundups. 9. Proptech press / editorial backlinks.

Sitelinks (the indented sub-links under EliseAI's result) are explicitly not a build target: Google generates them algorithmically and they cannot be requested or marked up. They are an earned byproduct of being #1 for the brand term + clean site architecture (unique titles, descriptive meta per page) + steady branded search volume — all of which Tracks A/B already advance.

Consequences

Commits us to:

• A measurement-first cadence: revisit keyword targeting only once GSC query data populates (~1 week); no keyword work on invented volumes.
• One-concern-per-PR discipline for Track A, held for per-PR approval (multi-PR session).
• Track B copy going through Sean/positioning sign-off before publish (customer-facing).
• An ongoing, non-engineering authority effort (Track C) that is the gating factor for AI-Overview and head-term visibility.

Honest expectations (so we don't over-promise):

• Branded terms: already winning ("propflow ai" = #1). Sitelinks and a knowledge panel follow as authority + branded volume grow.
• Long-tail / specific queries ("ai leasing assistant for small landlords", "automated maintenance triage software"): reachable in weeks with Track B.
• Head terms ("ai property management"): months, gated on Track B+C authority — not achievable by on-page or technical work alone.
• AI Overviews: months, gated on Track C third-party citations — not on schema or any markup.

Easier: brand disambiguation (entity schema), measurement (GA4/GSC/Bing), and a clear backlog the team can prioritize. Harder/slower: nothing is harder, but the authority track is long and largely outside engineering's direct control.

Alternatives considered

• Paid search to "boost ranking." Rejected as a ranking mechanism — Google Ads and organic are separate systems; ads do not feed the organic algorithm. Brand-defense ads on "propflow ai" (competitors V7/RealPage/Revela currently bid above our #1 organic) are a legitimate paid-media decision, but out of scope for this SEO ADR.
• llms.txt / AI-specific markup / content chunking for AI Overviews. Rejected — Google explicitly states these are not used for its generative features; effort better spent on indexable, well-structured, citable content.
• Chase head terms first with on-page optimization. Rejected — without domain authority, on-page tuning cannot rank a new domain for incumbent-owned head terms; sequencing long-tail + authority first is the realistic path.
• Rebrand to escape the name collision. Out of scope and disproportionate; entity schema + authority + consistent brand mentions are the proportionate fix.