# Depureco USA AI / LLM Website Database - v8 Final Wrap-Up

Generated: 2026-06-10

## Final recommendation

Use **v8** as the wrapped working package. Earlier versions are useful history, but v8 is the best current deployment/control package because it adds the final missing layers: product truth, model compatibility, feature-to-model logic, live page alignment, non-fit query guardrails, keyword refresh priorities, market trigger context, localization triage, and internal attribution cleanup notes.

## Public upload set

Upload these website files:

```text
/llms.txt
/llms-full.txt
/ai-sitemap.xml
/ai/*.md
/ai/db/*.json
```

Do **not** upload:

```text
/internal-review/*
```

## Technical SEO handling

Recommended handling remains:

```text
/llms.txt and /llms-full.txt: crawlable and indexable if desired
/ai-sitemap.xml: crawlable
/ai/*.md and /ai/db/*.json: crawlable with X-Robots-Tag: noindex, follow
/internal-review/*: do not upload publicly
```

## What v8 closes

1. Product claims now have a source/evidence map.
2. Model/filter/accessory compatibility has a safer quote-path framework.
3. Features such as Longopac, HEPA, PTFE, JetClean, antistatic, conductive accessories, pump-out, hoppers, and ORD LOC are mapped to product/category logic.
4. Page updates can be compared against GSC, execution queues, product copy, and SR-only copy needs.
5. Negative/non-fit intent is separated from public buyer education.
6. Keyword/opportunity refresh rows are consolidated for next SEO work.
7. Market/niche/process trigger logic is consolidated without exposing sales playbooks.
8. Spanish/localization pages are now triaged without making them a priority project.
9. UTM/source-quality cleanup is documented as internal-only.

## Stop point

At this stage, more database mining will likely produce diminishing returns until the files are uploaded and validated live. The next real work should be:

1. Upload v8 public files.
2. Confirm HTTP status for `/llms.txt`, `/llms-full.txt`, `/ai-sitemap.xml`, `/ai/*.md`, and `/ai/db/*.json`.
3. Confirm `X-Robots-Tag: noindex, follow` for `/ai/*.md` and `/ai/db/*.json`.
4. Test GSC URL Inspection on root files and a few `/ai/db/` files.
5. Monitor GSC, AI referral traffic, and Good Fit leads for 2-4 weeks.
6. Run the answer-engine QA rubric from v7/v8.
7. Rebuild only after live data shows what changed.