Superhuman vs Shortwave AI Search: Which Finds Things Faster?

This post compares one thing: AI search. Not who has more shortcuts, not who ships to Outlook, not who has the nicer split inbox. Just: which of Superhuman Mail or Shortwave finds things faster in a real mailbox, and where each one falls over.

That matters because search is the AI feature that most changes what an inbox is worth. If the archive answers questions in plain English, three years of mail becomes a knowledge base. If it does not, it stays a filing cabinet you file into and never open.

We build an email client too. It gets one short section near the end and no thumb on any of the scales above it.

The verdict, up front#

Shortwave wins the search dimension, and it is not close on the piece that decides most purchases: asking a question in natural language and getting an answer synthesised from across the mailbox and its attachments. That is the whole shape of Shortwave's product, and it is not the shape of Superhuman's.

Superhuman has built harder on speed and on the AI that sits beside the open thread than we have credit for; if your search job is mostly "summarise this thread", "draft a reply that references the last message" and "find the file so-and-so sent last week", Superhuman's assistant does that quickly and its keyboard shortcuts get you to the result in fewer keystrokes.

But for "what did we decide about pricing at the March offsite, and who was against it" — the shape of query a founder or an executive assistant actually types — Shortwave is the one that will surface the right threads and read them for you.

At a glance#

Both vendors move quickly and both restructure their AI features across tiers, so treat this table as the shape of the two products in August 2026, and verify the specifics on each vendor's own product and pricing page on the day you buy.

First: which Superhuman are we talking about?#

Half the comparison content on the web is wrong here, so worth thirty seconds. Grammarly acquired Superhuman, the email client, in July 2025. In October 2025 the parent company renamed itself Superhuman. "Superhuman" now covers two things: Superhuman Mail, the inbox, and the Superhuman Suite, a bundle around it.

This comparison is about Superhuman Mail. Anywhere below where it says "Superhuman", read "Mail". Anything that describes cross-app agents, docs or a bundled writing tool is describing the Suite and does not belong in a search comparison.

What "AI search" actually means in each product#

Both products do retrieval and both do generation, but they pointed those capabilities at different jobs. That is the whole comparison in one paragraph.

Shortwave's assistant is built retrieval-first. Its own product page describes it as "AI-powered search to find emails fast, or to analyze your entire team's emails and attachments for the answers you need". You ask a question in English, it searches the mailbox for the threads and attachments that plausibly hold the answer, then reads them and writes a synthesised reply with citations back to the messages. Saved searches and AI filters are extensions of the same primitive.

Superhuman's AI is built around the open thread. Its product page foregrounds an assistant that triages the split inbox, drafts replies, catches misses and executes short workflows. Structured search is fast and complete. Semantic search exists, and works well when you can point at a neighbourhood — this thread, this sender, this week — but it is not the shape of query the product is optimised around.

Both approaches are legitimate design choices. They are not the same product with different polish, and if you buy one expecting the other's search behaviour you will be disappointed.

Where Shortwave wins the search dimension#

Three jobs are markedly better on Shortwave, and if any of them is a job you do weekly, that is your answer.

• **Asking a question of the mailbox in natural language.** "Who was in the room when we agreed the Q3 pricing change, and what were the objections" is the shape of query Shortwave's assistant is built to answer. It searches, ranks, reads, and writes the answer with links back to the messages it used. That last part — citing the threads it read — is what makes the answer trustable rather than a plausible hallucination.
• **Sweeping across many threads at once.** Structured search will find every email from `finance@` in March. A semantic assistant will find every thread where money was actually discussed, whether the sender was `finance@` or `patricia@` or the invoice was pasted in as a PDF. The second is the search a human actually wants and the first is what old inboxes give them.
• **Analysing attachments as part of a query.** The product page names attachments explicitly. "What did the vendor quote for the Series B round" pulls the number out of the PDF rather than asking you to open eight of them.
• **Turning a repeated question into a saved search or AI filter.** Anything you ask more than twice can be pinned. This is what turns the archive into a live report rather than a lookup.

Where Superhuman wins the search dimension#

Ceded plainly: Superhuman has built harder on speed and on the AI beside the open thread than we or Shortwave have credit for.

Three jobs are faster on Superhuman, and if the shape of your search need is these three, do not let this post talk you out of it.

• **Getting to a message you can half-remember.** Command palette plus structured search plus the shortcut layer means you type three characters, arrow-down twice and press return before Shortwave's assistant has finished the first sentence of its answer. For a mailbox you know well, faster keystrokes beat cleverer retrieval.
• **Summarising the thread you are already in.** Superhuman's assistant is designed to work beside the current message. "What did they ask me" and "what is the state of this thread" are one keystroke and instantly useful.
• **Drafting a reply that references what was said.** Because the assistant already has the thread loaded, it drafts with that context without a separate query. Shortwave can do this too, but the round-trip is shorter on Superhuman.

Attachments and long threads#

Almost every real search question a working professional asks eventually hits an attachment. "The contract" is a PDF. "The deck" is a Google Slides link. "The invoice" is a spreadsheet. This is the axis where product intent shows most.

Shortwave publishes attachment analysis as part of the AI search surface. Ask about a number, a clause or a name and the assistant will open and read the relevant files as part of composing the answer, rather than surfacing them as blue links you then open yourself. That difference — read for you versus find for you — is the point.

Superhuman's approach is closer to "very good structured search plus a fast assistant on the open thread". Attachments are searchable. Reading a PDF as part of a mailbox-wide question is not the headline behaviour of its assistant.

On long threads specifically, both products summarise, and both do it well. The gap widens the moment the answer to your question spans several threads rather than one.

Indexing and privacy: what to check before you point either at your archive#

AI search means something is being indexed, and something is being sent to a model. Any comparison that skips this is skipping the part your security team will ask about.

Neither vendor publishes a full architecture diagram of what leaves the mailbox, when, and to which model. What each does publish is a security page, a privacy policy and, in Shortwave's case, documentation of what its assistant reads and how usage is metered. Read those on the vendor's own site rather than trusting a summary in a blog — including this one.

Two specific questions worth putting to sales in your trial:

• **What is indexed for AI search, and where does the index live?** Full-body plus attachments, headers only, or something in between; on their infrastructure or yours.
• **What is sent to a model at query time, and is any of it retained by the model provider?** "Zero retention with the model provider" is a specific contract term, not a marketing phrase, and you want it in writing.

Pricing model, not prices#

We are not going to quote a dollar figure for either product, because both restructured tiers within the past year and any number quoted in a blog is one release away from being wrong. What is structural, and what actually drives the cost of AI search, is the shape.

Both products are per-seat subscriptions. Both put AI features on the paid tiers rather than the free or trial one. Both meter AI use by tier rather than by a flat included allowance.

For Shortwave, the metering is specifically the thing that decides how much search you get: a daily allowance of AI requests, a cap on how many threads a single AI search will consider, a history window (how far back the assistant will look), and a limit on the number of AI filters. Each of those bites hardest on the person with the biggest archive — which is the person most likely to be shopping for AI search in the first place.

For Superhuman Mail, the tier decides whether Mail's AI is available at all and how deeply it is integrated. Because search is less of Superhuman's core positioning, tier movement is less about "how much search" and more about which AI features you get.

Model this on your real usage before you decide. Two hundred AI queries a week against a fifteen-year archive costs more than fifty queries a week against a two-year one, on either product.

Who each one is genuinely for on search#

Who should stop reading and buy something else#

If your mail is on Microsoft 365 or Exchange, Shortwave is not an option — its own docs say Microsoft 365 and Exchange are not supported, and sign-in requires a Google account. On search alone, that fact decides it for you: Superhuman Mail is the one to trial.

If your mail is on Fastmail, iCloud, Zoho or a self-hosted IMAP server, neither product is a clean fit for search on your real mailbox. Shortwave supports these only by auto-forwarding into a Gmail account, which means a copy of your mail sits in Google and search runs against that copy. Superhuman does not document IMAP support at all. If native IMAP is the constraint, the right search story is elsewhere in the category.

A third option, honestly#

Disclosure first: we build AI Emaily. This section is here because a comparison post that pretends the reader has only two choices is a worse post. Weigh what follows accordingly.

AI Emaily is a candidate on this dimension for one specific reader: someone whose mail lives across Gmail, Outlook and IMAP in one working day, and whose search question is "across all of it, what happened with X". We connect the three providers in one inbox and our semantic search — smart search plus ask-your-inbox — runs against that unified index rather than against one provider at a time. That is a shape neither Superhuman nor Shortwave targets, because both are pointed at one provider family.

Where each of the other two is ahead on search, they are ahead and we say so. Shortwave has built harder on Gmail-native semantic retrieval than we have; if your world is entirely Google Workspace and you want the deepest retrieval on that one provider, buy Shortwave. Superhuman has built harder on speed and on the assistant beside the open thread; if your search job is fundamentally "summarise this thread, draft the reply", buy Superhuman.

What we do add on top of search is control over what the assistant is allowed to do with what it finds. Copilot requires approval before any send, Autopilot is gated, every agent action is undoable, and each is written to an audit log. If your search question is "find the invoice and reply with the correct amount", the second half of that is where we do more than either.

The honest call#

On search across a Gmail or Google Workspace mailbox — the reader most comparison shopping for this — Shortwave is the pick. Its assistant does the retrieval-and-read job that changes what an archive is worth, and its attachment behaviour closes the loop most search stories break on.

On search across a Microsoft 365 mailbox, Superhuman Mail is the pick, because Shortwave is not available at all.

On search across a mixed mailbox — Gmail plus Outlook plus IMAP — trial ours. That is the specific gap we were built to close, and neither of the other two closes it.

Whichever way you land, run a real trial on a real mailbox in the same week. Two of your actual search questions decide this faster than any comparison, this one included.