ChatGPT Custom GPT Knowledge Files Not Being Used

Q: I edited the files but users still get old answers. Why?

You almost certainly didn't click **Update**. Configure preview uses your live draft; the shared GPT URL serves the last published version. Every Knowledge or Instructions change needs an explicit **Update**, then re-test on the public URL — not the preview.

Knowledge files configured but the GPT says "I don't see that file" mid-chat. Force retrieval, prune to under 12 files, and verify indexing — here is the full fix.

Published: May 21, 2026 Updated: Jun 17, 2026 Author: AI Productivity Guide Team 🌐 查看中文版本

Fastest fix: Custom GPT Knowledge files are not auto-injected into every turn. They go through on-demand retrieval (file_search), so the model only reads a file when the question matches the document’s wording, or when the Instructions explicitly order it to search first. Put one hard instruction at the top of Configure → Instructions — “Before answering, always search the knowledge files first” — click Update, and roughly 70% of “it ignores my files” cases clear up immediately.

Attaching a file in Configure is not the same as the model reading it every turn. The caps (20 files, 512MB per file, 2,000,000 tokens per file as of June 2026) are hard limits, but in practice the retrieval trigger rate is the real bottleneck, not the caps.

Which bucket are you in

Run this 30-second triage before touching settings. Open a fresh chat with the GPT and ask it to “quote one sentence verbatim from page 1 of <your-file>.”

What you see	Most likely cause	Jump to
GPT answers from general knowledge, never mentions files	Instructions don’t force a lookup	Cause 1
It works with exact document wording, fails on paraphrase	Question misses retrieval keywords	Cause 2
Some files quote fine, others always fail	Too many files, or one file too big	Cause 3
File is listed but every quote is wrong or refused	Silent indexing failure / file vanished	Cause 4
Knowledge holds only PNG / MP3 / MP4	Type accepted for upload, not for retrieval	Cause 5
Works in Configure preview, broken on the shared link	You never clicked Update	Cause 6

Common causes

Ordered by hit rate, highest first.

1. Instructions don’t explicitly require knowledge lookup

The most common failure. You upload PDFs / DOCX and assume the GPT will use them, but the default behavior is to answer from general knowledge first. Without an instruction like “always search the knowledge files before answering,” the model usually skips retrieval entirely.

How to spot it: Ask the GPT a fact that exists only in the knowledge files (e.g. an internal product codename). If it says “I don’t know” or fabricates a version number, retrieval did not fire.

2. The user question doesn’t hit retrieval keywords

Retrieval uses keyword / vector matching. If the question doesn’t contain words from the document, similarity drops below threshold and the chunk is skipped. Example: the file is “API rate-limit policy” and the user asks “why are my requests being blocked?” — “rate-limit” never appears, so the document isn’t retrieved.

How to spot it: Ask once using the document’s literal wording (“what is the API rate-limit policy”), then again with a paraphrase (“why are requests blocked”). If only the literal version works, that’s keyword drift.

3. File count near the 20 cap, or single file too large

Custom GPT limits as of June 2026: 20 files maximum, 512MB and 2,000,000 tokens per file, with CSV/spreadsheets effectively capped near 50MB. All text is split into chunks and only the top-scoring chunks are pulled per query, so a large file retrieves only a few chunks — and the answer may sit in a chunk that wasn’t pulled. The 2M-token ceiling is usually hit long before 512MB for text-dense PDFs, which silently truncates the file.

How to spot it: Open Configure → Knowledge and check file count and size. More than 15 files, or any single file over 50MB, is the first thing to investigate. Ask the GPT to list every filename in its Knowledge; if it can’t list all of them, files are being dropped.

4. Silent indexing failure (or the file vanished)

The upload UI showing “success” is not the same as content being indexed. Corrupt PDFs, exotic encodings, and scanned PDFs without a text layer all fail silently during backend indexing — the file appears in the list but search returns nothing. A separate, still-reported bug (active on the OpenAI developer forum through 2026) is files disappearing from the Knowledge list after a save or page refresh, so the GPT genuinely has nothing to read.

How to spot it: Ask the GPT to “quote one sentence verbatim from page 1 of xxx.pdf.” If it can’t quote, or the quote is fabricated, indexing failed. Then reload Configure → Knowledge and confirm the file is still listed at all — if the count dropped, it was silently removed and must be re-uploaded.

5. File type accepted for upload but not for retrieval

Images, audio, video can be attached to Knowledge but are not vectorized and indexed — they sit on disk. The model genuinely cannot see them when asked.

How to spot it: Knowledge contains only PNG / MP3 / MP4 and queries always fail — this is the cause. Convert to PDF / DOCX / TXT / MD.

6. Tested in Configure preview but published version is stale

When you test in Configure it uses the current draft. The GPT URL users see (or you use day-to-day) is the last version you hit Update on. Changes to files / instructions without pressing Update never reach the public URL.

How to spot it: Open the public GPT URL, ask “list files you can access,” compare to Configure → Knowledge. Mismatch means you forgot to Update.

Before you start

Confirm whether the issue happens in the Configure preview or only on the published GPT URL — they use different version snapshots.
Duplicate the current chat or screenshot it before retesting so history doesn’t pollute the next diagnostic.
Confirm your plan. Free accounts can use published Custom GPTs (within tighter message limits) but cannot build one — the GPT editor requires Plus ($20/mo), Pro, Team, or Enterprise as of June 2026. If you can’t open Configure at all, that’s a plan issue, not a file issue.

Info to collect

File type, size (MB), pages / rows; whether it’s a scanned PDF, whether it has Chinese / formulas / tables.
Total file count in Knowledge and whether near the 20 cap.
Full prompt text + GPT reply screenshot; record verbatim if it says “I don’t see the file.”
Full text of the current Configure → Instructions (screenshot it).
Whether you tested in Configure preview or via the published GPT URL.

Shortest fix path

Ordered by ROI. The first two usually solve 70% of cases.

Step 1: Add a hard requirement to the top of Instructions

Open Configure → Instructions and put this at the very top (not buried at the end):

Before answering ANY question, first call `search the knowledge files`
to check if the answer exists in the uploaded documents. Only use your
general knowledge if the knowledge files contain nothing relevant.
When citing knowledge, name the source file.

Use English even for non-English GPTs — instructions are followed more reliably in English. Click Update to publish a new version.

Step 2: Ask using keyword-style prompts

Train yourself (and your users) to use literal wording from the documents:

Bad:  How does this feature work?
Good: According to `user-guide.pdf`, how does "bulk export" work?

Bake the “always quote the source filename” pattern into instructions so the model also models it back to users.

Step 3: Prune Knowledge to ≤ 10 core files

20 is the hard cap, but 8-12 well-named files give the most reliable retrieval. Beyond 15 files, retrieval skews to the few highest-scoring ones and ignores the rest.

Steps:

Configure → Knowledge → delete stale / redundant files.
Merge several small files into one well-structured doc with clear headings (# Chapter 1) — chunks retrieve with more coherent surrounding context.
Give files semantic names: api-rate-limit-policy.pdf beats doc1.pdf.

Step 4: Split oversized files + convert to Markdown

For single PDFs > 50MB:

# Split PDF (macOS / Linux)
brew install qpdf
qpdf --split-pages=30 large.pdf split-page-%d.pdf

# Convert to Markdown (much higher extraction quality)
pip install marker-pdf
marker_single large.pdf ./output --max_pages 100

Uploading Markdown instead of PDF gives noticeably better retrieval, especially for table-heavy docs.

Step 5: Verify indexing succeeded

After publishing, open a brand-new chat and ask:

1. List every filename in your Knowledge.
2. Quote one sentence verbatim from page 1 of `<filename>.pdf`.
3. How many pages / rows is that document?

Any failure means that file’s index is broken — delete and re-upload to trigger re-indexing.

Step 6: Publish and test via the GPT URL

Don’t test only in Configure. Click Update → Share → copy the GPT URL, open it in an incognito window or a different account, and re-run the three checks from Step 5. Working in Configure but broken on the URL means you didn’t Update.

How to confirm the fix

Open a fresh chat (don’t continue in the old one), upload the same file, ask the same question — confirms it’s not chat history contamination.
Open the GPT URL in incognito or a different account and retest — rules out account / client caching.
Have a colleague repeat the flow in their own account — confirms it’s fixed for everyone, not just you.

If still broken

Cut the file to the absolute minimum: 100-row CSV, 1-page PDF — see if the smallest case works.
Swap formats: PDF → txt, DOCX → md, to rule out a format-specific parser issue.
Switch model: in Configure → Model, set the GPT to GPT-5.5 (the default since April 2026) rather than leaving it on an older default; a Thinking/Pro reasoning mode triggers retrieval more reliably than the Instant mode for ambiguous questions.
Check status.openai.com — a degraded “file search” or “GPTs” component will make every retrieval fail regardless of your config.
Package the source file, prompt, GPT URL, and subscription-level screenshot, then file a ticket at help.openai.com.

Prevention

When building a Custom GPT, prepare a 5-10 question “retrieval trigger test set” — facts that only exist in Knowledge. Run it before publishing every update.
Always put “first search the knowledge files” in the first sentence of Instructions, never buried mid-paragraph.
Keep files at 8-12 total, ≤ 30MB each, with semantic filenames.
After every Knowledge / Instructions change, hit Update and verify via the public GPT URL — not the Configure preview.
For tables / data, prefer CSV / Markdown over PDF — extraction-layer quality is dramatically better.

FAQ

Why does my Custom GPT ignore the Knowledge files but read a file I drop into the chat? A file you attach in the chat window is added directly to that turn’s context, so the model reads it immediately. Knowledge files are not in context — they sit behind file_search and are only pulled when the query matches or the Instructions force a search. That gap is exactly why “uploaded files work, Knowledge files don’t” is one of the most common reports.

How many Knowledge files should a Custom GPT have? The hard cap is 20, but retrieval quality is best at roughly 8–12 well-named files. Past about 15 files, retrieval skews toward the few highest-scoring documents and starves the rest. Fewer, larger, well-structured files beat many tiny ones.

Does adding “always search the knowledge files” really change anything? Yes, more than any other single change. The model’s default is to answer from general knowledge first; one explicit instruction at the very top of Configure → Instructions flips that default and is the highest-ROI fix. Write it in English even for a non-English GPT.

My file shows in the list but the GPT can’t quote it. Is it broken? Almost certainly an indexing failure — a corrupt PDF, a scanned PDF with no text layer, an unusual encoding, or the 2,000,000-token-per-file cap silently truncating a large document. Delete the file and re-upload to force re-indexing, or convert it to Markdown/TXT first.

I edited the files but users still get old answers. Why? You almost certainly didn’t click Update. Configure preview uses your live draft; the shared GPT URL serves the last published version. Every Knowledge or Instructions change needs an explicit Update, then re-test on the public URL — not the preview.

Can Free users use my Custom GPT? Yes. Free accounts can open and use published Custom GPTs (with tighter message limits), so retrieval should work for them too. What Free accounts cannot do is build a GPT — the editor needs Plus, Pro, Team, or Enterprise.

Tags: #ChatGPT #Troubleshooting