On this page
If Scholar will not open and Zotero will not sync, treat them as separate connections first
A common case is that Scholar can find a paper, but the publisher page does not provide the full text. Or an item appears in Zotero, while its PDF is missing on another device. This is not contradictory: Google Scholar search, DOI redirects, publisher full text, item sync, and attachment sync are separate connections. Find the earliest failing segment so the rest of the diagnosis stays on track.
Where the failure occurs
| Action | Service actually accessed | What to check first |
|---|---|---|
| Google Scholar search | scholar.google.com, Google sign-in, and verification | Whether the outbound route changes frequently and whether a human-verification challenge appears |
| DOI redirect and full text | doi.org, the publisher, and the institutional authentication portal | Final domain, account authorization, and institutional VPN |
| Zotero item sync | zotero.org sync service | Sync queue, account status, and whether the connection enters Clash |
| PDF attachment sync | Zotero Storage or your own WebDAV | Storage address, certificate, capacity, and sustained upload |
If Scholar search works but the publisher returns 403, do not assume one rule caused both outcomes. Likewise, if Zotero items sync but a PDF is still missing, Scholar is not the issue. Record the first failing URL and the time it occurred so the connection log has something concrete to search.
Create one Research policy group only for entry points you use long term
Research accounts and long downloads are poor candidates for an automatic group that frequently changes outbound routes. Create a manual selection group, place stable nodes first, and route Scholar, DOI, and core Zotero services through it. Add publisher and WebDAV domains based on your own traffic, because institutions and storage providers do not share a universal list.
proxy-groups:
- name: Research
type: select
proxies:
# 替换成当前配置中真实存在的节点名
- REPLACE_WITH_STABLE_NODE
- DIRECT
rules:
- DOMAIN,scholar.google.com,Research
- DOMAIN-SUFFIX,doi.org,Research
- DOMAIN-SUFFIX,zotero.org,Research
# 把自己的出版社或 WebDAV 域名加在这里
- MATCH,DIRECTREPLACE_WITH_STABLE_NODE is only a conspicuous placeholder. Replace it with a node name that already exists on the proxy screen before saving. MATCH,DIRECT lets unlisted traffic continue to connect directly. If your existing configuration already has a fallback rule, keep it instead of adding another one.
If Scholar repeatedly asks for verification, keep the browser session and outbound route unchanged
Human-verification prompts on Scholar search are often triggered by many requests in a short period, cookie state, or a changing outbound route. Stop refreshing and complete verification on the same node. An automatic benchmark group that changes routes every few minutes can make a newly verified session look suspicious again.
If a private window can search normally on a fixed node while your regular window remains stuck in a loop, first clear site data or extensions related to Scholar and Google sign-in. If the page explicitly shows an account, region, or access-frequency restriction, follow its instructions; Clash cannot replace account permissions.
If DOI redirects work but full text will not open, check whether TUN is intercepting institutional authentication
An institutional VPN, EZproxy, or single sign-on service often depends on internal DNS, private address ranges, or specific callbacks. First confirm that the institutional portal works with Clash TUN disabled, then enable the Clash system proxy and test the browser. If the system proxy already handles research access, there is no reason to let a second virtual adapter take over the institutional VPN.
If you must use TUN at the same time, exclude the institution's documented private address ranges and internal domains, and keep authentication callbacks on the same path. If a publisher says “not subscribed” or requests institutional sign-in, the issue is authorization; do not try to evade it by repeatedly changing regions.
403 after opening a DOI
Record the publisher's final URL, then check the account, institutional authorization, and current outbound route without rebuilding the Scholar rule.
Institutional sign-in page loops
Disable automatic node switching, then inspect callback domains and institutional VPN routing.
Campus domain does not resolve under TUN
Bypass Clash for internal DNS and private address ranges, or use only the system proxy.
Zotero item sync and attachment sync have two separate statuses
Official Zotero documentation separates library data from attached files: account sync handles items, tags, and notes, while PDF attachments use Zotero Storage or WebDAV. The 2026 documentation still states clearly that WebDAV applies only to personal libraries; group libraries cannot use WebDAV.
Create a test item containing only a title and click the toolbar sync button. If another device or zotero.org can see the item, data sync is complete. If the attachment still fails, open “Settings → Sync” to inspect the file-sync method, and use “Verify Server” for WebDAV instead of resetting the entire library.
Zotero sync result
| Result | Next step |
|---|---|
| Both items and attachments fail | Check the account, zotero.org requests, and system time |
| Items succeed, PDF fails | Check the Zotero Storage quota or the WebDAV address, certificate, and capacity |
| The browser downloads a PDF, but Zotero shows no connection record | Check whether Zotero uses the system proxy; use TUN or an application proxy if necessary |
| Desktop succeeds, mobile WebDAV fails | Confirm that the address uses HTTPS; mobile blocks plain HTTP by default |
Complete the final check with one paper
The final proof that your rules are correct is one real research workflow. Choose a paper you are authorized to access and follow it from the search result to the attachment on another device. This makes the broken step easier to identify than opening several test sites at once.
From search to another device
Search using a fixed Research node
Open one result and confirm in the connection log that scholar.google.com matched the intended group.
Follow the DOI to the final page
Record the publisher domain; when institutional access is required, use the legitimate portal provided by your institution.
Save to Zotero
Confirm that the browser Connector passes the item to the desktop app.
Sync manually
Watch the item queue and attachment queue separately; do not combine their outcomes.
Open the PDF on another device
The workflow is complete only when both the title and attachment appear, confirming that search, save, and sync all succeeded.