Written and maintained by CASRAI Editorial Board
Last updated
A PubMed saved search is not the same thing as a PubMed alert, and treating them as one setting is why most researchers’ alerts stop being useful within a few months. A saved search is a stored query. An alert is a separate, optional layer on top of that query: a scheduled job that re-runs it and emails you what changed. PubMed also offers a third option that skips the account entirely — an RSS feed of the same query. This guide covers how to set up each one through My NCBI, how the email schedule actually works, and — the part most walkthroughs skip — the query-maintenance habits that keep an alert from either flooding your inbox or going silent for good.
Saved search, email alert, and RSS feed are three different things
PubMed’s search page exposes two separate links under the search box once you run a query: Create alert and Create RSS. They both start from the same saved query, but they behave differently in ways that matter for which one you should use:
| Email alert | RSS feed | |
|---|---|---|
| Requires signing in to My NCBI | Yes | No |
| Delivery | Email, on a schedule you set | Pulled by your feed reader whenever it checks |
| Where the query lives | Tied to your NCBI account | Encoded in the feed URL itself |
| Can expire from inactivity | Yes — account-level, see below | Not documented by NCBI either way |
| Item volume control | You see everything new since last run | You set a display limit, up to 200 items |
The practical dividing line: use the email alert if you want PubMed to actively interrupt you on a schedule. Use the RSS feed if you already read a feed reader daily and would rather pull updates on your own terms — or if you don’t want a query tied to a login at all, for example on a shared or institutional search you’re setting up for someone else.
Setting up an email alert through My NCBI
The email alert is the most common setup and the one most institutional guides walk through:
- Run your search in PubMed as you normally would, refining it until the result set is one you’d actually want repeated. (See the section below on why this step matters more than it looks.)
- Click Create alert, directly under the search box.
- If you aren’t already signed in, PubMed will prompt you to sign in to (or register for) an NCBI account — this is required for email alerts specifically; it is not required for the RSS option.
- Set a name for the saved search and choose your email schedule (see the next section), then save.
Once saved, the search appears in the Saved Searches section of your My NCBI dashboard, where you can re-run it manually at any time, see how many new results have accumulated since the last run, or edit it.
Choosing (and later changing) the alert frequency
My NCBI’s documentation specifies three schedule options for automatic email updates: daily, weekly, or monthly. There’s no single right choice — it depends on how the underlying query behaves:
- Daily makes sense for a narrow, well-specified query in a fast-moving area (a specific gene-disease pairing during active investigation, a named trial’s ongoing publication output) where you genuinely want to see something the day it’s indexed.
- Weekly is the reasonable default for most standing literature-monitoring searches — a research topic, a methods area, a competitor’s lab output — because PubMed indexing has enough day-to-day noise that a daily digest is often either empty or a single record.
- Monthly suits broad background-awareness searches you want to keep running for years without maintenance overhead: a whole subfield, a standing “watch this journal” query, anything you’re tracking for context rather than urgency.
You are not locked into your original choice. To change the schedule later, go to the Saved Searches list in My NCBI and click the gear icon next to the search you want to edit — that opens the same configuration screen you used to create it, where the schedule (and the underlying query itself) can both be changed. Save to apply.
Why the query you save matters more than the schedule
This is the part that determines whether an alert is still useful in month six, and it’s mostly ignored by setup instructions that stop at “click Create alert.” A saved search inherits every property of the query you ran it from, including properties you didn’t think about at the time:
- Free-text terms drift. A query built on keywords rather than MeSH terms will miss newly published papers that use different terminology for the same concept, and will pick up unrelated papers as terminology shifts in the field. Building the alert around the relevant MeSH heading, with subheadings scoped deliberately, is more stable over a multi-year alert than a purely keyword-based query — at the cost of missing very recent papers that haven’t been indexed yet, since MeSH indexing lags publication.
- Over-broad queries decay into noise, not silence. Unlike a query that’s too narrow (which just returns nothing and is easy to notice and fix), an over-broad alert keeps arriving on schedule while slowly becoming less relevant as the field around your topic grows — the failure is gradual and easy to stop reading rather than obviously broken.
- Filters you added once can silently exclude new relevant work. A publication-date range, article-type filter, or language restriction added when you first built the search stays baked into the alert indefinitely unless you go back and edit it. It’s worth reviewing an alert’s underlying query every few months against a search you build fresh, to see whether it’s drifted from what you’d actually search for today.
- A query built with the Advanced Search Builder is easier to audit later than one typed as a single free-text string, because the field tags and combined history are visible and editable rather than buried in one long string you have to re-parse to understand.
A useful habit: schedule a recurring personal reminder — not a PubMed feature, just a calendar note — to re-run each standing alert’s query manually every 3–6 months and compare it against a fresh search on the same topic. That single check catches most of the drift above before it costs you a genuinely relevant paper.
The RSS alternative in more detail
To create an RSS feed instead of (or alongside) an email alert, run your search and click Create RSS under the search box. PubMed generates a feed name (editable, defaults to your search terms), lets you choose how many items the feed displays via a dropdown, and produces a feed link once you confirm. You can manually edit the limit= parameter in the resulting URL to display up to 200 items — useful for a broad topic where you want your feed reader to hold more history — though a higher limit increases the feed’s load time.
Because the RSS option doesn’t require an NCBI account, it’s a reasonable choice when you want a search’s ongoing results available to whoever reads your feed reader without tying it to a personal login, or when you’d rather pull updates on your own schedule than have PubMed push them to your inbox. The tradeoff is that PubMed’s own documentation does not specify a retention or expiration policy for the feed link itself, unlike the account-level expiration that applies to saved searches (below) — treat a long-lived RSS feed the same way you’d treat any other saved query: worth an occasional manual check that it still resolves and still reflects the query you intended.
Managing, re-running, and deleting saved searches
From the Saved Searches section of My NCBI, each entry shows a What’s New column with a hyperlinked count of new results since the search last ran — a way to check for updates without waiting for the next scheduled email, or without an alert configured at all. Clicking the search name re-runs it immediately.
To remove a search, go to Manage Saved Searches, check the box next to the one you want gone, and click Delete selected item(s). There’s no NCBI-documented cap on how many searches an account can hold, so there’s little reason to leave stale ones around rather than deleting them when a project ends.
One account-level detail worth knowing if you set up an alert and don’t touch it for a long stretch: My NCBI accounts with no activity for two years are deleted, which takes every saved search and alert tied to that account with them. For an alert you intend to run for years with minimal maintenance, that’s a real (if distant) reason the periodic manual check described above is worth keeping up, rather than treating “set it and forget it” literally.
How this compares to a Google Scholar alert
If you’re deciding between the two rather than running both, the mechanics differ in ways that matter for reliability: PubMed’s alert is built on a controlled, taggable query against a single well-defined biomedical database, with a documented three-tier schedule and an RSS escape hatch that needs no account. Google Scholar’s alert system covers a much broader, cross-disciplinary index but is built on an undocumented query parser with a 256-character limit and no published schedule options, which makes it harder to build an alert you can predict the long-term behavior of. For biomedical literature specifically, a well-built PubMed alert is generally the more auditable and durable of the two; Scholar’s broader coverage matters more for interdisciplinary or grey-literature-adjacent topics PubMed doesn’t index at all.
Frequently asked questions
Do I need an NCBI account to save a search in PubMed?
You need to sign in to (or register for) an NCBI account to create an email alert. You do not need an account to create an RSS feed of a search — that option is available without signing in.
How often can a PubMed alert email me?
Three schedules are available: daily, weekly, or monthly. You can change the schedule at any time from the Saved Searches section of My NCBI by clicking the gear icon next to the search.
Why did my PubMed alert stop finding anything new?
Usually query drift rather than a broken feature: a keyword-based query that no longer matches how the field describes the topic, a date-range or filter left over from when the search was first built, or a topic that’s genuinely been fully covered by earlier alert emails. Re-run the underlying query manually and compare it to a fresh search on the same topic to check.
Can I export a PubMed RSS feed’s results directly into a citation manager?
The RSS feed itself is a standard feed URL, so any feed reader can subscribe to it; whether a particular reference manager can consume that feed directly depends on the manager, and isn’t something PubMed’s own documentation addresses — check your reference manager’s own feed-import support before relying on it as an automatic pipeline.
How many saved searches can I have in My NCBI?
NCBI’s documentation does not state a maximum number of saved searches per account.
What happens to my saved searches if I stop using my NCBI account?
NCBI accounts with no activity for two years are deleted, which removes any saved searches and alerts tied to that account along with them.
For building the query worth saving in the first place, see the PubMed Advanced Search Builder and MeSH terms in PubMed. For a broader look at literature-discovery tooling beyond PubMed itself, see the research tools hub.
