A sourced research brief in minutes

It's late afternoon and someone needs "a quick note on where solid-state batteries are" before tomorrow's planning meeting. A chatbot will give you five confident paragraphs in seconds, with no way to tell which lines came from a journal and which from a press release. The alternative is an evening of trade press and paywalled abstracts. This page shows what a CiteJury research brief gives you instead: what's established, what's contested and what's still open, each line tied to a source, and how to read a brief that comes back Contested.

Why a summary needs sources now

Fluent summaries fail in ways that are hard to spot. In a study coordinated by the EBU and led by the BBC, journalists from 22 public service media organizations evaluated more than 3,000 answers from ChatGPT, Copilot, Gemini and Perplexity. 45% had at least one significant issue, and 31% had serious sourcing problems: missing, misleading or incorrect attributions.

The cost of passing those errors along is going up. In May 2026, arXiv said that authors whose submissions show incontrovertible evidence of unchecked AI output, such as fabricated references, face a one-year ban, after which their papers must first be accepted at a peer-reviewed venue. You don't have to publish research for the lesson to apply. If a line in your brief can't be traced, it can't be defended.

What goes into a CiteJury brief

Choose Research a topic and type the topic the way you'd say it to a colleague, with a date or a region if it matters. CiteJury searches several independent search indexes and favors primary sources: journals, regulators, government statistics, filings and official pages. It looks for evidence that cuts against the obvious story as well as for it, and it reads long PDFs and reports, not just the first screen of a page.

The best passages go into one sealed evidence pack with an ID, and nothing can be added later. Claude, ChatGPT and Grok each write a brief from that pack, separately and without web access, and every quote they cite is checked word for word against its source. A final review writes the brief you read: a few short sections with every line linked to a source, a call on how settled the topic is, and a list of what the evidence leaves open. The three outcomes are Well established, Contested and Thin evidence.

Below is the finished result of a real brief, shortened to fit.

“The state of solid-state batteries in 2026”

Contested

Claude and Grok agree. ChatGPT said Thin evidence18 sources usedTook 10 minutes

One thing is settled: semi-solid cells ship today, all-solid-state cells are still pilots.

• What's established

  Semi-solid batteries are in production in 2026.

• Who leads

  Company claims conflict and are mostly self-reported.

• Performance and cost figures

  Not independently verified in the sources.

Safe way to say it

As of mid-2026, trade sources agree semi-solid batteries are in production while all-solid-state cells remain at pilot and prototype stage.

01grandpasai.comKey evidence

02theweeklydriver.comBackground

03nature.comUnreliable

Reading the solid-state batteries brief

We asked for "The state of solid-state batteries in 2026" on Standard. The brief drew on 18 sources and took 10 minutes, longer than most Standard checks. It came back Contested. Claude and Grok agreed on that call, and ChatGPT said Thin evidence.

Only one point was settled: semi-solid batteries are in production in 2026, while all-solid-state cells are still at the pilot and prototype stage. Everything else split. On who leads, company claims conflict and are mostly self-reported. On performance and cost, the figures weren't independently verified in any of the sources. So the brief tells you what you can say today, and it's narrower than most headlines.

The safe way to say it was: "As of mid-2026, trade sources agree semi-solid batteries are in production while all-solid-state cells remain at pilot and prototype stage." Notice the attribution. The settled point rests on trade sources, and the wording says so, which lets your reader weigh it.

The source rulings are worth a look too. A page on grandpasai.com was marked Key evidence, theweeklydriver.com was Background, and a page on nature.com was marked Unreliable. That last ruling is about that page and this question, not the journal's name. A famous masthead doesn't make a page evidence for a specific 2026 claim, and a less famous site can carry the one line that holds up. The rulings tell you which sources to open first.

The split between Contested and Thin evidence helps rather than confuses. Both readings tell a planner the same thing: don't quote a market leader or an energy-density figure from this evidence yet.

How to get a brief you can use

1. Scope the topic. Add a date and a boundary, such as "in 2026", "in the EU" or "in adults over 65". A broad topic gets a broad brief.
2. Put the angle in the topic. "Solid-state batteries in 2026: who ships, and what independent tests show" keeps the brief on the points you need.
3. Pick a tier. Standard, about 8 credits, suits most topics. Deep, about 32 credits, suits fast-moving or niche topics, because the AIs also run their own web searches while sources are gathered. Quick, about 2 credits, uses one AI with no final review and is fine for getting your bearings.
4. Read the outcome, then the open questions. They tell you what not to say before you decide what to say.
5. Open the sources behind any line you'll reuse. Check the ruling next to each one.
6. Ask follow-ups. They answer only from the same sources and cost about 1 to 2 credits, so the brief doesn't drift.
7. Keep the trail. Share the link, or download the sources as Excel or CSV to keep with your notes.

Habits of careful researchers

• Separate "exists" from "at scale". "In production" and "in mass production" are different claims. The battery brief settled only the first.
• Label self-reported numbers. A company's range or cost figure is a claim until someone independent measures it. Write "the company says".
• Go to the most direct source. A journal paper beats a news story about it, and the news story beats a blog post about the news story.
• Date the brief. Write "as of" with a month. A brief on a moving field has a short shelf life.
• Treat Contested as a finding. Present both sides with their sources rather than picking the one you prefer.
• Use the open questions as your agenda. They're the best list of what to ask an expert or read next.
• Cite what you opened. For coursework or a paper, the brief is a map. Cite the original source, not the brief. There's more on this in checking the facts in your paper before you submit.

What it won't do

It isn't a systematic review. A brief rests on the sources the search found, 18 in the battery example, not on every paper ever published. Results can be wrong or incomplete, which is why every line links to its source.

Many journal articles sit behind paywalls, and CiteJury can't read pages behind paywalls or logins. It lists the pages it couldn't open, so you know which ones to get through your library. Scanned PDFs without a text layer can't be read either, since there's no OCR.

It checks text, not charts, figures or video, and in technical fields a lot of the data lives in figures. Very recent developments may have little coverage yet, which usually means Thin evidence. It doesn't format or validate a bibliography, and it isn't medical, legal or financial advice. For topics where the rules matter, checking what a regulation actually says goes straight to the official text.

A good brief is short and honest about its edges. The battery example settled one point and said so plainly, which is more useful than five paragraphs that sound settled and aren't. When you need a single answer rather than a map, ask a question and get an answer you can check.

Questions people ask