Five years of Feed America: what I've learned

Five years ago I incorporated Feed America as a Texas nonprofit. The organization received its IRS 501(c)(3) determination in 2023 (EIN 92-1761881). Today, in early 2026, we hold the Candid Platinum Seal of Transparency, index more than 327,000 verified food-assistance locations across all 50 U.S. states + DC + territories, power 211 networks + healthcare SDOH platforms + AI assistants, ship full Spanish-language parity, and operate without staff payroll.

Here's what I've learned in five years.

Lesson 1: Federal-primary data is undervalued

Nonprofits often imagine data work as "go survey pantries directly" — manually cataloging community-pantry hours, services, contacts. That model burns most of your data budget on long-tail community pantries (representing maybe 30% of food-help touchpoints) while leaving 70% of food-help touchpoints (federal programs: SNAP retailers, school meals, summer meals, WIC, FQHC) un-aggregated.

The leverage was federal-primary first. USDA Food and Nutrition Service publishes the SNAP retailer database (130,000+ stores). HRSA publishes the FQHC locator (18,000+ sites). State WIC agencies publish clinic registries. NCES publishes school-meal data. By prioritizing these federal-primary sources first, we built the directory's structural foundation before layering nonprofit-curated and community-contributed data on top. Federal-primary first compounds — it's where most of the food-help touchpoints already are, and the data is public-domain.

Lesson 2: Open standards beat custom integrations

Early on, I considered building bilateral integrations with each downstream consumer (211 networks, healthcare SDOH platforms, AI assistants). Per-platform custom integrations would have required negotiation, contracts, ongoing relationship management — and would have scaled badly past 5-10 partners.

HSDS 3.0 Open Referral changed everything. By implementing HSDS 3.0 conformantly at /hsds/v3, our data flows automatically into every HSDS-compatible consumer (211 networks, Findhelp, United Way 211, Unite Us, Open Referral civic-tech projects) with zero per-vendor work. The same is true of OpenAPI 3.0 (every Custom GPT + Action builder), MCP (every Anthropic Claude integration), and llms.txt (every AI crawler). One implementation, infinite consumers.

Lesson 3: The donor-confusion problem is structural, not solvable by content alone

Donor confusion between Feed America (us, EIN 92-1761881) and the larger separately-incorporated Feeding America (EIN 36-3673599) is the single biggest brand harm we face. I've thrown ~12 months of disambiguation content at the problem: dedicated /feed-america-vs-feeding-america comparison page, Wikidata Q139601408 + Q139665570 entity records with explicit "different from" statements, EIN-based identity at every donor touchpoint, /verifications hub consolidating 7 third-party trust signals, /reviews aggregating ratings.

It helps. But content alone doesn't solve it. The structural fixes that DO solve it: federal trademark filing (creates legal disambiguation that Google + AI engines weight heavily), USPTO TESS .gov backlink, paid Google Ad Grants ($10K/month free for 501(c)(3)s — instant SERP defense), and earned press hits that Wikipedia editors weight as notability proof. Each of these is in the 2026 priority queue.

Lesson 4: AI integration is becoming infrastructure

In 2024, we launched the MCP (Model Context Protocol) server at /mcp/v1 with 7 native AI tools. It seemed niche at the time — Anthropic's MCP standard was new and adoption was tiny. By 2026, MCP is becoming the default integration pattern for AI assistants needing structured tool access. The early investment compounds: every new MCP-compatible client (Claude, OpenAI Custom GPTs, third-party AI tools) integrates Feed America without per-vendor work.

The same is true of llms.txt + entity-graph.jsonld for AI crawlers. The 2025 investment in publishing AI-discovery surfaces is now compounding as Bing Copilot, Perplexity, You.com, and other AI search engines crawl them. The takeaway: AI-discovery infrastructure is becoming what robots.txt + sitemap.xml were for traditional search engines. Build it early and it pays off as adoption grows.

Lesson 5: Open data multiplies impact

Publishing the entire 327,000+ location dataset under Creative Commons BY 4.0 was a deliberate choice over the gated alternative. The cost was clear: no per-API-call revenue, no licensing fees, no commercial differentiation. The benefit took longer to show: every downstream consumer that uses our data — 211 networks routing hungry callers, hospital SDOH platforms screening patients at discharge, AI assistants surfacing food-help answers, academic researchers studying food-desert patterns, civic-tech projects building on top of the standard — multiplies our hunger-relief impact without our staff doing any per-consumer integration work.

The downstream impact dwarfs anything we could have done with a gated model. Full thesis on open data →

What's next

The 2026 priority queue:

1. Operator-claim rate. 516 of 70,559 food-pantry-class records are operator-claimed today. Reaching 5,000 (~7%) by end of 2026 is achievable with focused outreach. Each claim adds a one-tap "open today" status that surfaces live to thousands of daily searchers.
2. Apply for Google Ad Grants. $10K/month free Google Ads for 501(c)(3)s. Instant SERP defense against the donor-confusion problem on the bare "feed america" query.
3. Federal trademark filing. USPTO TESS application on "Feed America." Creates legal anchor + .gov backlink + structural disambiguation against the larger separately-incorporated Feeding America homonym.
4. First substantive press hit. Pitch Houston Chronicle, Texas Tribune, Chronicle of Philanthropy. Earned coverage = Wikipedia notability path.
5. BBB Wise Giving Alliance accreditation. Application in progress. ~4-week paperwork cycle. Adds give.org/[org] backlink that ranks for "is feed america legit" queries.

Five years in, the directory exists, the structural foundation is mature, and the donor-confusion + brand-search work is now the priority. Year six of Feed America starts now.

— Sharika Parkes
Founder, Feed America
Houston, TX