Compare them by what you are left holding.

Six questions worth asking of any tool in this category, answered against each company's own published pages. On four of the six, something else is the better answer — and those four come first. The two that are left are the reason this is being built.

Join the practitioner waitlist

Titrate is pre-launch. What is described here is what it is being built to do, with Jeff Casebolt, PhD.

Topic stack6 papers · 2015–2025

The central question

Does the barbell hip thrust produce superior gluteus maximus development — measured either as acute EMG activation or as chronic hypertrophy — compared to the back squat?

Contreras et al.2015n=13supports
Delgado et al.2019n=8mixed
Plotkin et al.2023n=34contradicts
Krause Neto et al.2025n=12mixed
Vigotsky et al.2017n=—contradicts
Kassiano et al.2024n=33supports

222

2 supports2 contradicts2 mixed

The comparison this category would rather you ran.

Every tool here will tell you how many papers it has. The claims run from 100 million to 314 million, several vendors contradict themselves between two of their own pages, and not one of the numbers carries a date. It is a contest about the size of a borrowed thing: the one corpus in the set that is demonstrably an upstream rather than a resale is Semantic Scholar's, it is free, and Consensus names it as one of its own sources. Running that comparison tells you almost nothing about what you will be holding at the end of a Tuesday afternoon. So this page runs a different one — starting with the table you came for, which does not settle it either.

The table, drawn the way the category draws it

Four steps. Who does which.

Find the literature. Keep it somewhere it stays. Draft something cited from it. Publish that under your own name. Every product below was checked against its own published pages on 2 August 2026, and every mark means one of five specific things.

What each mark means

Yes: Stated on the company's own published page.
Partial: Present, but narrower than the column.
No: Explicitly absent. An answer we found, not a blank.
Not claimed: Not claimed on any page we could fetch. Absence of evidence, not evidence of absence.
Specified: Titrate only. Committed in the schema and not shipped — a different kind of claim from every other mark on this table, printed differently on purpose.

Four steps, by product16 products · 4 steps

Product

Finds the literature

Keeps a library

Drafts cited output

Publishes under your name

Research tools

Elicit

Yes

Finds the literatureYes: Its own index, 138 million papers by its own count. The full operator grammar — +by:, +journal:, +year: — is Pro and above.
Keeps a libraryYes: Extraction tables and a systematic-review workflow.
Drafts cited outputYes: Research Reports: a structured report with citations, up to 80 papers.
Publishes under your nameNo: No occurrence of publish, byline, your name or audience anywhere in reachable copy.

Consensus

Yes

Partial

Not claimed

Finds the literatureYes: Names its own upstreams outright: Semantic Scholar, OpenAlex, its own crawl, PubMed.
Keeps a libraryPartial: A Library you can save papers into.
Drafts cited outputNot claimed: No report or writing feature found on any page we fetched.
Publishes under your nameNo: Zero occurrences, re-checked live.

scite

Yes

Not claimed

Finds the literatureYes: Crossref-anchored and DOI-gated, with full Elasticsearch query syntax — the most rigorous query grammar of the six.
Keeps a libraryYes: Collections, capped at 1,000 papers on Basic and 10,000 on Pro.
Drafts cited outputNot claimed: The Assistant grounds its answers in papers; no drafting product was found.
Publishes under your nameNo: Zero occurrences.

ResearchRabbit

Yes

Finds the literatureYes: A different paradigm: citation-graph discovery seeded by articles, not keyword narrowing. The four-keyword workflow does not map onto it.
Keeps a libraryYes: Collections, and the graph persists.
Drafts cited outputNo: Its own Write better tagline links to exploration tools, not to composition.
Publishes under your nameNo: Zero occurrences. Author appears only in the sense of a paper's authors.

Semantic Scholar

Yes

Not claimed

Finds the literatureYes: The only one of the six with true author-entity disambiguation — claimable author pages, not string matching. Free, with no paid tier at all.
Keeps a libraryNot claimed: Not claimed on any page we fetched.
Drafts cited outputNot claimed: Not claimed on any page we fetched.
Publishes under your nameNo: Zero occurrences.

SciSpace

Yes

Not claimed

Yes

Finds the literatureYes: Over 200 million papers from various sources — the only one of the six that names no upstream at all.
Keeps a libraryNot claimed: Not claimed on any page we fetched.
Drafts cited outputYes: An AI Writer and a Generate Verified Report feature.
Publishes under your nameNo: Zero occurrences.

Reference managers

Zotero

Yes

Finds the literatureNo: It does not discover; you bring papers to it.
Keeps a libraryYes: Its core function, and better at it than anything else on this page.
Drafts cited outputNo: No drafting. As of Zotero 9 in April 2026 there are no AI features in it at all.
Publishes under your nameNo: Zero occurrences.

Paperpile

Yes

Partial

Finds the literatureYes: Through a Google Scholar integration.
Keeps a libraryYes: Its core function.
Drafts cited outputPartial: Citation and bibliography insertion into Docs and Word — not drafting.
Publishes under your nameNo: Zero occurrences.

Find, organise, author — then stop

Gemini Notebook

Yes

Finds the literatureYes: Discover Sources searches the open web and returns up to ten annotated recommendations.
Keeps a libraryYes: Per-notebook source panels.
Drafts cited outputYes: Reports, including a Blog Post format, every one traceable to its sources.
Publishes under your nameNo: The word throughout is share, never publish. No author name is shown to a viewer, and there is no list.

Logically

Yes

Finds the literatureYes: Research, Cite, and Write With AI — its own tagline.
Keeps a libraryYes: A reference manager, with import from the others.
Drafts cited outputYes: Document Writer, with citations attached as you write and 10,000+ citation styles.
Publishes under your nameNo: Zero occurrences of publish, byline, your name or audience.

Undermind

Yes

Partial

Finds the literatureYes: Deep literature search across databases, following citation trails.
Keeps a libraryYes: Workspaces and paper libraries.
Drafts cited outputPartial: Custom tables and reports with in-line citations.
Publishes under your nameNo: Zero occurrences.

STORM (Stanford)

Yes

Finds the literatureYes: Multi-engine retrieval.
Keeps a libraryYes: A pre-writing outline stage.
Drafts cited outputYes: Writes full articles with citations from a topic.
Publishes under your nameNo: Its own README: the system cannot produce publication-ready articles that often require a significant number of edits.

Closes all four

Perplexity Pages

Yes

Partial

Yes

Finds the literatureYes: Deep Research across the general web — no academic-literature mode comparable to a PubMed or Semantic Scholar index.
Keeps a libraryPartial: Threads and Spaces. Ad hoc, not a persistent structured library.
Drafts cited outputYes: Cited synthesis, described in its own copy as usable as a work product.
Publishes under your nameYes: In its own words: Publish your work to our growing library of user-generated content and share it directly with your audience with a single click. Featured pages carry a by [Name] byline.

Publishing platforms

Substack, Ghost, beehiiv, Medium

Yes

Finds the literatureNo: No research features of any kind.
Keeps a libraryNo: No citation management. None of the four has a reference library.
Drafts cited outputNo: No research grounding. Ghost and Medium mention no AI writing at all.
Publishes under your nameYes: Your name, your list, your domain. This half of the market is clean, and so is the other half — the two do not currently meet.

Research reviews

MASS, Stronger by Science, Examine, Barbell Medicine, Renaissance Periodization

Yes

Finds the literatureYes: An editorial team does it.
Keeps a libraryYes: An editorial team does it.
Drafts cited outputYes: An editorial team does it, monthly, and has for years.
Publishes under your nameNo: You are the reader. The byline belongs to the brand, and so does the authority it earns.

Titrate

Specified

Finds the literatureSpecified: Topic Finder is specified and its output contract is committed — a paper record already carries its source type, authors, journal, year and DOI. The retrieval is not built: there is no search endpoint in the repository.
Keeps a librarySpecified: Articles, and topics that a paper is read against. Committed in the schema.
Drafts cited outputSpecified: A position on a question, carrying its citations on the record. Not a generated paragraph — you write it.
Publishes under your nameSpecified: Content pieces, digests and subscribers exist as schema. Nothing has been published through it, because nothing has launched.

Checked 2 August 2026 against each company's own published pages, or its own README where that is the primary source. Where a page could not be fetched, the cell says not claimed rather than guessing.

Read this table and Titrate is not the only thing that closes all four. Perplexity Pages closes them in its own words. The research reviews have closed them for years, with an editorial team doing the work and the brand's name on the front — that is not a gap in the market, it is a business, and it is a good one. So the table does not settle it, and a version of this page that pretended otherwise would be exactly the thing you came here to check.

No screen is depicted here and no product is depicted here — this is a table of what each company says on its own site, and the citation for every cell is that company's page on the date given. Titrate's row is marked Specified throughout, and that mark means something weaker than every other mark in the table: committed in the schema, not shipped. Nothing has been published through Titrate, because nothing has launched.

The six questions

What actually decides it.

These are the questions that change what you are holding when you are finished. Four of them have a better answer than Titrate. Those four are first, and the reason they are first is that a comparison page written by the product it favours is worth reading only for the parts where it does not.

Where do the papers come from, and does the tool tell you?

What it decides

Whether you can audit what you were shown, or only trust it.

Titrate loses this one

Semantic Scholar: The best answer in the set, and free. It is an upstream corpus rather than a reseller — built by the Allen Institute for AI from publisher partnerships, data providers and web crawls — and Consensus names it as one of its own sources.
scite: The clearest provenance statement of the paid six: metadata from Crossref as a canonical source, only publications with a definitive DOI indexed, full text from 35 named publisher agreements.
Consensus: Names its upstreams outright — Semantic Scholar, OpenAlex, its own crawl, the entirety of PubMed, and direct full-text partnerships with six of the twelve largest publishers.
ResearchRabbit, SciSpace: Three data providers, unnamed. Various sources, unnamed. Those are the whole of what either says.
The numbers themselves: Consensus says 220 million on its help centre and 200 million on its own blog. ResearchRabbit says 310 million on its home page and 270 million in its guide. scite gives 100 million, 300 million and 314 million on three pages without reconciling them. None of the six carries a date.

Titrate

Titrate has no index. It would query the same public corpora the paid tools resell, so there is no version of this question it wins, and there is no corpus number on this site — not on this page, not anywhere. That is a permanent position on the current architecture, not a gap waiting to be closed.

Can you actually open what it finds?

What it decides

Whether a result is a paper you can read or a pointer to one you cannot.

Nobody wins this one

Consensus: Has genuine licences with six of the twelve largest publishers — Wiley, Taylor & Francis, Sage, APA — and states it uses that full text only to measure search relevance, and generate short, cited previews. It does not hand you the PDF.
scite: Has 35 publisher indexing agreements and needs the full text to extract citation statements. It does not redistribute it either.
The publishers: Elsevier permits text mining by researchers at subscribing institutions, for non-commercial research, with outputs capped at snippets of up to 200 characters. That is the ceiling every product in this category is working under.
PubMed Central: Indexes 11,327,616 articles. About 3.4 million of them — roughly 30% — are in the open-access subset that may be redistributed. The rest are free to read and conventionally copyrighted.

Titrate

Titrate will not hand you paywalled full text either, and it is not going to pretend the constraint is a design choice. The one thing available here that nobody does is arithmetic: state what proportion of a topic you can actually open before you search it, rather than after. The next block is that number for one real question.

Does what you found stay?

What it decides

Whether next month's version of you can reconstruct why you kept a paper.

Titrate loses this one

Zotero: Twenty years old, free at 300MB, $20 to $120 a year for more, and as of Zotero 9 in April 2026 it still has no AI features in it at all. If storing and syncing references is the job, this is the answer, and a new product should say so.
Paperpile: $4.15 to $5.75 a month, annual billing only, with full-text PDF search and Docs and Word integration.
The research tools: scite's Collections cap at 1,000 papers on Basic. ResearchRabbit's caps are on seed articles — 50 free, 300 paid — not on what you keep. Consensus has a Library. None of them is trying to replace a reference manager.

Titrate

Titrate is not competing on storage and would lose if it did. What it adds is the field a reference manager has no place for: the question the paper was read against. A folder of PDFs is not a position, and twenty years of folders is still not a position.

What does it cost — and what does doing it without any of them cost?

What it decides

What you are actually paying for, once the free floor is on the table.

The floor is zero

Free: PubMed. Google Scholar. Semantic Scholar, which has no paid tier at all. ResearchRabbit's core. Zotero at 300MB. Substack and Medium, free to publish on. PubMed plus Zotero plus Substack is $0, and that is the honest price this is competing against.
Research tools: Elicit Pro $49/mo billed $588/yr, Scale $169/mo billed $2,028/yr. scite $20/mo and $50/mo. SciSpace $12, $70 and $160/mo on annual billing. Undermind $16/mo. Humata $9.99/mo.
Reference managers: Paperpile $4.15 to $5.75/mo, annual only. Zotero storage $20, $60 or $120 a year.
Research reviews: MASS $29/mo or $299/yr. Barbell Medicine $19.95/mo or $199/yr. Renaissance Periodization $24.99/mo billed $299.99/yr.
Could not be verified: Consensus and Examine. Both are bot-walled and rate-limit even a proxy; three passes failed on each. The $19 Examine figure that circulates is an illustrative grandfathering example on Examine's own help page, not a live price. We are not going to print a number we could not fetch.

Titrate

Titrate has no price. It is pre-launch, the waitlist is the only thing to join, and pricing is not decided. When it is, it will be on a page like this one with the same dating on it. Everything above was captured on 2 August 2026 from each vendor's own pricing page; prices move, and these will be wrong before they are old.

Where does your reading of the evidence live?

What it decides

Whether you own a position you can defend, or a folder you have to re-read.

Nothing in the set answers this

Reference managers: Store papers. They hold no position. That is not a criticism — it is what they are.
The research tools: Return a result set, or a report you read once. Elicit's Research Reports, Consensus's Deep Search, SciSpace's Verified Report: each is a document that ends when you close it.
scite and Consensus: Do label how the literature leans — Smart Citations for supported or contradicted by later research, and the Consensus Meter for Yes, No, Possibly and Mixed. Both operate on a search result, not on a position you wrote and put your name to. Titrate does not claim that labelling as its difference; see the concessions below.

Titrate

A position in Titrate belongs to a topic, and its citations live on the record itself: an identifier pointing at the article in your own library, plus the formatted reference, per claim. The commitment is that the trail is part of the record rather than a habit you maintain in a second app — not that the field can never be empty, which would be a stronger claim than the schema makes.

When you publish it, whose name and whose list?

What it decides

Whose authority the work builds. This is the one that made the product.

One named exception, and it splits three ways

The six research tools: Zero occurrences of publish, byline, your name or audience anywhere in their reachable copy. Re-verified live on 2 August 2026 across all six. This is not an oversight on their part — an internal research tool has no reason to carry that vocabulary.
Perplexity Pages: The exception, and it is named here rather than left for you to find. Its own copy: Publish your work to our growing library of user-generated content and share it directly with your audience with a single click. Featured pages carry a by [Name] byline. It closes the loop.
What separates it: Three things, each checkable. Its corpus is the general web, not an academic-literature index. Its organise step is Threads and Spaces, not a persistent structured library. And you publish onto their domain, into their library, with no subscriber list of your own — which is the same problem as reading someone else's research review, moved one layer up.
Substack, Ghost, beehiiv, Medium: Your name, your list, your domain — and zero citation management or research grounding, verified across all four. The two halves of this market do not meet today.
The research reviews: MASS, Stronger by Science, Examine, Barbell Medicine and Renaissance Periodization already run find, organise, author and publish as a going concern. It works. The byline is the brand's, and the authority accrues to the brand.

Titrate

This is the whole reason to build it: the piece carries your name, it goes to a list that is yours, and it is built from a library you assembled. All three are separate schema commitments, and it is the conjunction rather than any single clause that nothing else in the set matches.

Every competitor claim in these six answers was fetched from that company's own published page on 2 August 2026 and is quoted or paraphrased close to its own wording. Two prices could not be verified and say so. Titrate's own answers describe a pre-launch product: the schema is real and checkable, the running software is not.

The one question above that resolves to a number.

One question, on PubMed

Papers indexed on the question
269100%
Free to read somewhere
9736%
In the redistributable open-access subset
3613%

Question 02 is the only one of the six with arithmetic in it, so here is the arithmetic, for one contested question — does passing objective return-to-sport criteria reduce second-ACL-injury risk. Sixty-four percent of that literature can only ever reach you as a citation, and no product in this category changes that, because none of them redistributes paywalled full text. What none of them does either — checked across all six — is tell you the proportion before you search rather than after. That is not a capability and it is not hard. It is a sentence, it costs nothing, and its absence everywhere else is the most informative thing on this page about how these products are built. The person building this is on the wrong side of it too: three of Jeff Casebolt's five peer-reviewed papers are behind a publisher paywall — Unpaywall returns closed for all three, checked 3 August 2026. He cannot hand you those either.

The queries these counts came from

Stated plainly, because you were going to find them

Four places this loses — and is not going to stop losing.

A comparison page that concedes nothing is an advertisement. These are the four positions where something else on this page is the better answer, what it loses to, and what is being done instead of competing there.

The corpus

Loses to

PubMed and Semantic Scholar. Free.

PubMed does five of the six search parameters this product was specified around — subject headings, author field, journal field, date ranges, a result-count control — and it is free. Semantic Scholar does them best, including the only true author disambiguation in the set, and it has no paid tier. Titrate has no index and would be querying the same public corpora. There is no corpus number anywhere on this site and there is not going to be one.

Instead

Compete on where the result lands, not on the search that produced it.

The reference manager

Loses to

Zotero. Twenty years of it.

Free at the core, no AI features as of April 2026, and better at storing and syncing references than anything being built now. If a reference manager is what the job needs, that is the answer and it costs nothing.

Instead

Add the thing a reference manager has no field for — the question a paper was read against — and do not pretend to replace the rest.

Labelling supporting, contradicting and mixed evidence

Loses to

Consensus and scite. Already shipped.

The Consensus Meter reports Yes, No, Possibly and Mixed. scite's Smart Citations show whether a finding has been supported or contradicted by later research. Both shipped before this page existed. And in Titrate today there is no stance column in the schema at all — it is a design commitment on a pre-launch product, not a feature, and it is not what makes the difference. Anywhere else on this site that a stack shows how a paper bears on a question, that is what it means.

Instead

Say so, here, rather than let a reader who already uses either tool find the overlap on their own.

What ships today

Loses to

All of them.

Elicit, Consensus, scite, ResearchRabbit, Semantic Scholar and SciSpace work right now. Titrate does not. The precise state, so it can be checked rather than taken: the destination is built — a paper record already carries its source type, authors, journal, year and DOI, and there is an index on the year column — and the retrieval is not, because there is no search endpoint in the repository. That sentence is checkable against the code. A coming soon badge would not be.

Instead

Publish the state, keep it accurate, and let the waitlist be the only thing asked for until it changes.

None of these four is scheduled to be fixed. Three of them are structural: no index, no intention of rebuilding a reference manager, and a labelling mechanism two named competitors already ship. The fourth is a matter of time — but until it changes, the sentence that describes it stays on the page.

Questions

Why not just use PubMed, Zotero and Substack? That combination is free.

For a lot of people that is the right answer, and it is the honest competitive price this is measured against. What it costs instead is continuity. The search happens in one place, the papers land in a second, the position you formed while reading them exists nowhere at all, and the piece you write in a third has to reconstruct the citations by hand. Every seam is a place where the trail back to the paper breaks — and the trail breaking six weeks later, when someone challenges the claim, is the specific problem this exists for. If your three tools already hold together, keep them.

Why not Elicit or Consensus? They find papers better than Titrate will.

They do, and that is not being contested. Elicit has its own index of 138 million papers and a real operator grammar; Consensus names its upstreams and has licensed full text from six of the twelve largest publishers. Both are better at finding than anything Titrate is being built to do. The difference is what you are holding afterwards: with either of them you have a report you read, and the authority for anything you then write is yours to reassemble by hand. Neither product carries the words publish, byline, your name or audience anywhere in its copy — checked again on 2 August 2026 — because neither is trying to.

Perplexity Pages already publishes under my name. What is actually different?

Three things, all checkable. First, its corpus is the general web; there is no academic-literature mode comparable to a PubMed or Semantic Scholar index. Second, its organise step is Threads and Spaces — ad hoc, not a persistent structured library you keep adding to over years. Third, and the one that matters most: you publish onto Perplexity's domain, into Perplexity's library, with no subscriber list of your own. That is the same problem as resharing someone else's research review, one layer up. Perplexity Pages is the honest exception to the claim on this site, which is why it is named rather than omitted.

Which of these is actually best at the search parameters Titrate was specified around?

Semantic Scholar, and it is free. Of keyword refinement, year range, author, journal, a defined number of results and a citation fallback when there is no PDF, it does five of six well and it is the only tool in the set with true author-entity disambiguation — claimable author pages rather than exact string matching. Elicit also does five of six, but its operator grammar is limited to Pro and above. PubMed does five of six for nothing. Consensus has no author filter among its eleven filter categories at all, and SciSpace states plainly that there is no dedicated author filter. If the search itself is the job, the answer is a free tool.

Why is there no price for Titrate on a page full of everyone else's?

Because it is not decided, and inventing one to fill the column would be the same category of thing as inventing a testimonial. Titrate is pre-launch; the waitlist is the only thing to join. The competitor prices are here because they are published, checkable and dated — and because the most useful number on the page is the floor, which is $0. When Titrate has a price it will be published with the same dating on it.

Everything here is dated 2 August 2026. What happens when it changes?

It goes out of date, and some of it already will have. Prices move, features ship, and two of the figures on this page could not be verified at all — Consensus's and Examine's, both bot-walled through three passes. Every claim carries the company's own page as its source and a capture date, so a stale cell is visible as stale rather than passing as current. If you find one that has moved, it has moved; check the vendor, not this page.

If questions 05 and 06 are the ones that matter to you.

Join the practitioner waitlist

One email when the first practitioner cohort opens. Nothing else.

The mechanism, in fullThe queries behind question 02, the four-keyword funnel that collapses to one wrong paper, and the four steps this page only names.

The hallucination objectionThe other thing you were going to check. A real topic stack where zero of seven papers support the claim, and what a summariser does with that.

Who is building itFive peer-reviewed publications with their DOIs, so the credential is checkable at the point of reading rather than asserted.