# Philippine AI Report — Practitioner Survey 2025 (open data)

> Anonymized results from the survey behind the 2025 Philippine AI Report: **179 respondents**, **41 questions**, fielded **June–December 2025**. Free to use under **CC BY 4.0**. Query it as JSON or through an **MCP server** — no API key, no signup.

**Canonical page:** https://philippineaireport.com/survey.md
**Structured record (Schema.org Dataset):** https://philippineaireport.com/api/survey.json
**Agent hub:** https://philippineaireport.com/for-agents

## Quick answers

- **What is this?** Microdata and aggregates from the practitioner survey that produced the report's headline findings (92% of organizations report active AI adoption, 65% stuck at pilot, 57% blocked by talent scarcity).
- **Who answered?** 179 practitioners and leaders at Philippine organizations — job level, AI budget authority, industry, company size, in-house vs third-party AI tools, GPU access, hiring, benefits, challenges, satisfaction.
- **How do I get it?** REST JSON endpoints (below) or MCP tools. Both are public and CORS-open.
- **Is it safe to publish analyses from it?** Yes. Respondent keys are irreversible salted pseudonyms; identifying questions never reached the database; small cells (< 2 respondents) are suppressed; quotes are redacted and human-approved.
- **How do I cite it?** "The 2025 Philippine AI Report — Practitioner Survey. Philippine AI Report / Swarm, 2026. https://philippineaireport.com/survey.md"

## Access via MCP (for AI agents)

Endpoint (Streamable HTTP, no auth):

```
https://swtwqrwsyedfamedawbm.supabase.co/functions/v1/mcp
```

Claude Code:

```sh
claude mcp add --transport http philippine-ai-report https://swtwqrwsyedfamedawbm.supabase.co/functions/v1/mcp
```

Claude Desktop / Cursor / Windsurf (`mcpServers` config):

```json
{ "mcpServers": { "philippine-ai-report": { "url": "https://swtwqrwsyedfamedawbm.supabase.co/functions/v1/mcp", "transport": "http" } } }
```

Survey tools:

| Tool | What it returns |
|---|---|
| `get_survey_meta` | Title, fielding window, `n_collected`, `n_published` |
| `list_survey_questions` | All 41 questions, types, answer choices, and `external_id` keys |
| `get_survey_aggregate` | Counts and percentages per answer option; optional `question` |
| `get_survey_responses` | One record per respondent with all answers; `limit` (≤1000), `offset`, optional `question` |
| `get_survey_quotes` | Redacted, human-approved open-ended answers; optional `question` |

Recommended flow: call `list_survey_questions` once, pick `external_id` values, then call `get_survey_aggregate` for headline numbers or `get_survey_responses` for cross-tabs.

Full tool manifest (all 14 tools): https://philippineaireport.com/api/mcp.json

## Access via REST (JSON)

Base: `https://swtwqrwsyedfamedawbm.supabase.co/functions/v1/survey-api`

| Path | Description |
|---|---|
| `/` | Endpoint directory, privacy guarantees, licence |
| `/meta` | Survey metadata and response counts |
| `/questions` | Question inventory with types and choices |
| `/aggregate` | Per-choice counts and percentages. Optional `?question=<external_id>` |
| `/responses` | Row-level anonymized records. `?limit=` (max 1000) `&offset=`, optional `?question=` |
| `/quotes` | Approved, redacted free-text answers. Optional `?question=` |

Example:

```sh
curl "https://swtwqrwsyedfamedawbm.supabase.co/functions/v1/survey-api/aggregate?question=250805112"
```

Percentages are computed against the number of people who answered that question, not the survey total.

## What the survey asked

Firmographics (job level, budget authority, who leads AI strategy, industry, organization type, headcount) · AI deployment (in-house vs third-party tools, usage patterns, most-used tools, tenure with generative AI) · Talent (AI roles employed, headcount, AI-related layoffs) · Investment (last-year and next-year AI budget, current and forecast GPU compute) · Usage and outlook (past-year and next-year use cases, what would improve adoption) · Outcomes (benefits, challenges, satisfaction, success over 12 months, whether AI caused headcount changes) · Tool evaluation (accuracy, ease of use, security, vendor support, integration, customization) · Personal (paying for own AI tools, monthly spend, willingness to help build internal tools, and why).

The full list with answer choices is at `/questions` or via `list_survey_questions`.

## Privacy and method

- Instrument: SurveyMonkey survey 523922236; self-administered online questionnaire.
- Questions collecting names, emails, employer, referrers, payment details, or free-text job titles were excluded at ingest and are absent from the database.
- Respondent keys are salted, irreversible pseudonyms, stable across releases.
- Aggregates merge answer options chosen by fewer than 2 respondents into a single "Other / suppressed" row per question (`min_cell` is returned in every aggregate response).
- Row-level records satisfy k-anonymity; free text is never included at row level.
- Quotes are machine-redacted and individually approved by a human reviewer; they are not linked to respondent keys.

## Licence

CC BY 4.0. Attribute "The 2025 Philippine AI Report" and link to https://philippineaireport.com.

## Related

- Report markdown: /report.md · Report metadata: /api/report-meta.json
- Discovery index: /llms.txt · Full-text bundle: /llms-full.txt
- Source code (MCP + API): https://github.com/internetoftim/ph-ai-hub-directory (`supabase/functions/mcp`, `supabase/functions/survey-api`)