This AI Does Your ENTIRE Systematic Literature Review in Minutes  🔗 **Source:** [This AI Does Your ENTIRE Systematic Literature Review in Minutes](https://www.youtube.com/watch?v=CRIYKjTOJA8) > [!note]+ Document Notes > tags: [ai-research-tools, systematic-review, literature-review, prisma-reporting, academic-writing] > author: Andy Stapleton > source: Andy Stapleton (YouTube) > reference: https://www.youtube.com/watch?v=CRIYKjTOJA8 > date created: 2026-09-29 10:32 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > # 🏷️ Title: This AI Does Your ENTIRE Systematic Literature Review in Minutes (PRISMA Included) > <!-- deep-read:sections start --> ## 🧭 Table of contents - [📝 Summary](#-summary) - [📖 Main content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning styles assessment](#-learning-styles-assessment) - [📊 Accessibility assessment](#-accessibility-assessment) - [🌍 DEIB assessment](#-deib-assessment) - [✅ Logical validity](#-logical-validity) - [🛡️ DARVO analysis](#️-darvo-analysis) - [🎯 Conclusion](#-conclusion) - [🔄 Refactor appendix](#-refactor-appendix--this-ai-does-your-entire-systematic-literature-review-in-minutes-prisma-included) ## 📝 Summary - **What happened.** Andy Stapleton's channel published a 12:59 walkthrough on 2026-09-29 of AnswerThis, an AI research tool. Stapleton tests two jobs: a "literature review" run on flexible transparent composite electrodes, and a "systematic review" run that follows the PRISMA reporting standard. The auto-transcript renders the product as "answer this," Zotero and Mendeley as "Zotterero" and "Mendlay," and arXiv as "archive." Quotes below keep the transcript spelling. - **The claim.** The narrator says this is one of the best literature and systematic review tools they've tested. They point to four features: academic search filters, generated figures with source highlighting, an export-friendly writer, and a PRISMA diagram that updates live. The closing verdict: "I cannot see a better way of doing it right now" (00:12:14). - **The mechanism.** For the literature review you set sources, date range, minimum citations, journal quartile, length and citation style, then click go. For the systematic review you enter a research question, the tool proposes PICO eligibility criteria and search strings, then walks you through fetching, screening, extraction and risk of bias. Reviews can be blinded to a human co-reviewer by email invite. - **Numbers shown.** The PRISMA panel reads 240 papers found and 203 after duplicate removal at the screening stage shown. The finished spaced-repetition review for undergraduates lists 29 included studies. - **Why it matters.** Systematic reviews take months. A tool that produces auditable search strings, screening logs and a PRISMA flow diagram shifts the work from writing the review to checking it. - **Key stakeholders.** AnswerThis (vendor), Andy Stapleton (creator, who sells their own courses and newsletter, see References), and academics, students and research librarians who run reviews. - **Risks and tradeoffs.** The narrator says twice that the user must check every step and that "auto run all steps" is unfit for anything submitted. The narrator shows no accuracy test: no recall or precision against a human-screened set, and no check that citations exist. The video shows one system-generated review and does not compare it with a human review. No sponsorship is disclosed in the transcript or description; the description carries only the creator's own commercial links. - **Bottom line.** The workflow design (editable PICO, reproducible search strings, live PRISMA counts, human second reviewer) is the strong part. The claim that it's the best available rests on one demonstration and the narrator's impression. ## 📖 Main content **The core argument.** AnswerThis bundles many tools (quick Q&A, presentations, data analysis, research gaps, find papers, chat with paper, paraphraser, trial landscape, per the 00:00:00 list) into one subscription. The narrator focuses on two: literature review and systematic review. The narrator's thesis is that the literature review has the most detailed academic filters they've seen in an AI tool, and that the systematic review flow stays auditable instead of behaving like "a black box." **Literature review settings (00:00:41 to 00:02:51).** Source options include papers you upload yourself (useful if you already keep a Zotero or Mendeley library), Semantic Scholar, open access papers, medical sources, trial registries, regulatory sources and arXiv, plus the open web. Advanced options: citation double-check (verifies the citation exists and supports the claim), publication date range, document type, minimum citation count, and journal quality at Q1 or Q2 and above. Output options: a structure with topics you specify, brief, standard or detailed length (the narrator picks 3,000+ words), academic tone, and a citation style. A "normal" versus "max" toggle sets depth; the narrator recommends max on the paid tier. **The output (00:02:51 to 00:04:16).** The demo review covers high-efficiency flexible transparent composite electrodes (the narrator briefly types "OPV devices" as an example topic first). Clicking a claim shows its source, and where a PDF exists the tool highlights the passage. The tool also generated a chart titled "maximum reported PCE" (likely power conversion efficiency, the standard photovoltaic metric; the transcript doesn't expand the acronym) and exposes figures for download. The narrator calls it the first review tool they know of that creates its own figures. Whether the chart values match the cited papers is not checked on camera. **AI Writer (00:04:16 to 00:05:43).** Edits happen in a WYSIWYG editor with a separate AI Writer that offers citation reordering, an AI detector and a plagiarism detector. Export goes to Word, PDF, Markdown or LaTeX, so users can leave the platform. Co-author invites and sharing are mentioned. **Systematic review flow (00:05:43 to 00:10:47).** You create a reusable workspace, saved to a default project, and enter a research question (the narrator picks a built-in asthma example). The tool drafts eligibility criteria as PICO: population (adults with asthma, including allergic asthma), intervention, comparator (standard care) and outcome, with include and exclude lists. Review mode can be AI alone, you plus AI, or you plus another person invited by email, which the narrator describes as blinded. "Form search string" then produces database search strings you can regenerate or hand-edit. Later steps are fetch papers, title and abstract screening, full text screening, data extraction and risk of bias. A single "auto run all steps" button skips manual gates. The narrator admits the interface doesn't always make the next click obvious (00:09:20). **PRISMA panel (00:10:04).** A flow diagram updates as papers move through stages. The screen shows 240 records identified and 203 after duplicates, with later stages still empty because the narrator hadn't run them. **Finished review (00:10:47 to 00:12:59).** The example topic is spaced repetition interventions for undergraduate students. The final document has abstract, background, methods, a generated PRISMA flow diagram, 29 included studies, a risk-of-bias table with a key, a list of studies, a discussion, a summary of findings, strengths and limitations, and a comparison with prior reviews. **Contested or unverifiable in this source.** - "Best" and "most impressive" are impressions from one run. No comparison with Elicit, Covidence, Rayyan or a human review appears. - The narrator says the systematic review is "PRISMA compliant" by repeating the tool's own label ("apparently this is Prisma compliant"). PRISMA is a reporting checklist, so a flow diagram alone doesn't establish compliance. - Screening accuracy, risk-of-bias validity and citation-check performance are not tested on camera. - The 3,000-word figure is a setting, not a measured output length. ## 🗒️ Notes #### ⏱️ [00:00:00](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=0) AnswerThis overview: literature review, systematic review and more - Opens on the landing page: quick Q&A, presentations, data analysis, research gaps, find papers, chat with paper, paraphraser, trial landscape. - The narrator picks literature review and systematic review as the two features worth testing. - Name garble: "answer this" is AnswerThis. > [!quote] [00:00:41] > I think this is probably one of the best systematic and literature reviews that I've seen any tool do at the moment. #### ⏱️ [00:00:59](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=59) Literature review settings: sources, citation checks, journal quality filters - Source scope: abstracts, open access PDFs, full text from your own uploads, Semantic Scholar, medical sources, trial registries, regulatory sources, arXiv, open web. - Citation double-check is switched on so cited papers exist and support the claim. - Filters: last 5 years, document type, minimum citations, Q1 or Q2 and above. - Length and style: brief, standard or detailed (the narrator picks 3,000+ words), academic tone, citation style, optional topics to cover. - Normal versus max: the narrator recommends max if you've paid for it. > [!quote] [00:00:59] > this is one of the most in-depth and detailed and academic kind of specific search filters I've ever seen in an AI tool. #### ⏱️ [00:03:13](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=193) Results: an AI literature review with its own generated figures - Pre-run example on transparent electrodes; sections include abstract, introduction and performance. - Clicking a claim shows the source and highlights the passage inside the PDF when one exists. - The generated PCE chart is a PNG created from the answer text; figures are downloadable. - Prompt advice: spend more time on the topic prompt than the narrator's quick "OPV devices" example. > [!quote] [00:02:51] > this is the first literature review that I can think of that actually creates its own figures. #### ⏱️ [00:04:30](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=270) AI Writer: edit, check plagiarism, export to Word, PDF or LaTeX - Recommends moving from the review view to a new document in AI Writer. - Tools named: reorder citations, AI detector, plagiarism detector, formatting options. - Export formats: PDF, Word, Markdown, LaTeX. The narrator jokes about how to pronounce LaTeX. - Collaboration: invite people, share documents. > [!quote] [00:04:16] > They are listening to academics. They are working with academics #### ⏱️ [00:06:06](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=366) Systematic reviews with AI: PICO eligibility criteria - Notes that systematic reviews "can take months" and should take time; the tool streamlines but the narrator says the process stays auditable. - New run creates a reusable workspace; the tool is labeled PRISMA compliant. - Enters a built-in research question on asthma; the review opens at eligibility criteria. - PICO fields shown: population (adults with asthma), intervention, comparator (standard care), outcome. > [!quote] [00:06:26] > it's not just a black box of AI doing stuff and giving you a response. This is completely auditable at the end #### ⏱️ [00:07:30](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=450) Refining search strings and screening papers - Three review modes: AI, you plus AI, or you plus a person invited by email. - Reviewer must check each AI-suggested criterion; then "form search string." - Strings can be regenerated by AI or hand-edited; "live editing" is praised. - Auto-run exists for demos. The narrator says not to use it for submitted work. - Steps listed at 00:09:20: fetch papers, title and abstract screening, full text screening, data extraction, risk of bias. The narrator concedes it isn't always obvious where to click. > [!quote] [00:08:37] > Obviously, you shouldn't do that if you're looking to submit this anywhere. #### ⏱️ [00:10:08](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=608) PRISMA diagram that updates as you go - Shows 240 papers found and 203 after duplicates; later boxes are empty because screening hasn't run. - The diagram fills in as papers pass each stage, so you can check counts per stage. - The narrator contrasts this with building PRISMA diagrams by hand at the end using other tools. > [!quote] [00:10:04] > it will update this as you go along so you can actually check in to see how many papers you're taking through each stage #### ⏱️ [00:11:19](https://www.youtube.com/watch?v=CRIYKjTOJA8&t=679) The output: a full systematic literature review with risk of bias - Example topic: spaced repetition interventions for undergraduate students; 29 studies included. - Document sections: abstract, background, methods, PRISMA flow diagram, risk of bias with key, included studies, discussion, summary of findings, strengths and limitations, comparison with prior reviews. - Closing claim: a human should still go through it with someone else and check everything, but for generating one easily the narrator sees no better option. > [!quote] [00:11:31] > This would have taken ages and obviously with a systematic review you go through it with someone else and you check everything ## 📚 References The transcript itself contains no links; the following come from the video description or are primary sources for named entities. - [This AI Does Your ENTIRE Systematic Literature Review in Minutes (PRISMA Included) — Andy Stapleton](https://www.youtube.com/watch?v=CRIYKjTOJA8) — source video, 12:59, uploaded 2026-09-29. - [Academia Insider newsletter](https://academiainsider.com/newsletter/) — in the description; "30,000+ email subscribers" is the creator's own claim. - [AI writing course](https://academy.academiainsider.com/courses/ai-writing-course) — the creator's own paid course, linked in the description. This is a commercial interest of the creator, though it isn't tied to AnswerThis. - [New course and community waitlist](https://www.subscribepage.com/andystapletonnewcourse) — the creator's own waitlist, in the description. - AnswerThis — the tool under review. The description doesn't link it, and no affiliate or sponsor link appears in the description or transcript. Find the official site and pricing before citing it. - [PRISMA statement](https://www.prisma-statement.org) — the reporting guideline and flow diagram standard the tool's systematic review mode targets. - PICO framework (population, intervention, comparison, outcome) — the clinical question format used for eligibility criteria; named at 00:07:11. - [Semantic Scholar](https://www.semanticscholar.org), [arXiv](https://arxiv.org), [Zotero](https://www.zotero.org), [Mendeley](https://www.mendeley.com) — named sources and reference managers; the transcript renders them "Semantic scholar," "archive," "Zotterero" and "Mendlay." - [LaTeX project](https://www.latex-project.org) — export format; the narrator asks viewers how to pronounce it. ## 🗣️ Paraphrase Stapleton argues that AnswerThis does two review jobs better than any tool they've tried. The literature review side starts with filters an academic would want: which databases, how recent, how many citations, which journal quartile, and whether citations exist. It then writes 3,000-plus words in which every claim links to its source, highlighted in the PDF, and it draws its own charts from the data it found. The AI Writer then lets you reorder citations, run AI and plagiarism checks and export to Word, PDF, Markdown or LaTeX, so you aren't stuck in the platform. The systematic review side follows the standard method in order. You state a question, the tool proposes PICO criteria you can edit, then it builds search strings you can inspect and change. It moves through fetching, screening, extraction and risk of bias, and a PRISMA diagram fills in as papers advance. You can run it with an AI, with an AI and yourself, or with a human second reviewer. Stapleton is careful about one point: the auto-run button exists so you can see what the tool does, not so you can submit the result. Everything else in Stapleton's pitch, including "best" and "cannot see a better way," is Stapleton's impression from a topic on spaced repetition that ended with 29 included studies. ## 📰 Newsletter **A PRISMA diagram that fills itself in.** Andy Stapleton walks through AnswerThis, an AI research tool that runs both narrative literature reviews and PICO-driven systematic reviews. The parts worth stealing for your own workflow aren't the AI writing. They're the design choices: editable PICO criteria instead of a blank prompt, database search strings you can read and change, a second-reviewer invite by email, and live PRISMA counts (240 found, 203 after duplicates in Stapleton's run) so you can see where records drop out. Where this lands for WordPress: contributor research on accessibility barriers, plugin security patterns and support-ticket themes all need reproducible search methods. The tool is a candidate for scoping such reviews. Treat the output as a draft for human screening, as the video says, because no accuracy test appears. Also check that the citation checker works before trusting a reference list. Links: [the video](https://www.youtube.com/watch?v=CRIYKjTOJA8), [PRISMA statement](https://www.prisma-statement.org), [Zotero](https://www.zotero.org). Disclosure: the creator sells their own courses; the description lists no AnswerThis link. ## ✅ Action items #### DevRel - Developer advocacy: if you cite this video in a talk, say it's a single-run demo by a creator with course products, and show the PRISMA counts, not the "best tool" line. - Education: run a session on reading a PRISMA flow diagram, using the 240-to-203 duplicate step as the first example. - Developer experience: docs for any AI-assisted research tool should show the editable-criteria pattern (PICO plus regenerate or hand-edit search strings) as a quickstart. - Developer marketing: don't copy "cannot see a better way" into material. Cite the mechanism (auditable search strings, live PRISMA counts) instead. - Community management: add a note in link-sharing channels that this is a creator review with no reported accuracy test. - Internal/external: ask AnswerThis for its citation-verification method and any screening recall figures before recommending it. #### OSPO - Policy and governance: decide whether AI-screened literature reviews are acceptable evidence for policy documents, and require a human second reviewer. - License compliance: check the tool's terms on uploaded PDFs (the "your own uploads" source at 00:00:59) and on reuse of generated figures. - Security management: uploaded papers and reference libraries are data flows to a third party. Add that to the review of any tool an OSPO recommends. - Software inventory: a hosted research tool is a network dependency with no export guarantee beyond the four formats shown (Word, PDF, Markdown, LaTeX). - Developer support: publish a checklist of what to verify by hand: five random citations, three screening decisions, and the PICO fields. - Community engagement: no direct action from this source; nearest relevance is funding evidence reviews for upstream projects. - Strategic alignment: set a KPI for review time per completed review with the human check included, so speed claims are measured. #### Community - Maintainers: use PICO-style scoping (who, what change, compared to what, which outcome) to write issue-triage research questions. - Contributors: pair a newcomer with a reviewer on a screening pass so the second-reviewer pattern from 00:07:54 becomes a contribution path. - Sponsors: if a sponsor funds an AI-assisted review, request the search strings and PRISMA counts as deliverables. - Ecosystem: no direct action from this source; nearest relevance is a shared protocol for cross-project reviews. #### WordPress - Core: no direct action from this source; nearest relevance is evidence reviews behind performance budgets. - Core Performance: no direct action; nearest relevance is a systematic review of published Core Web Vitals studies, screened by a human. - Core AI: add "AI-assisted evidence review" to guardrail docs with a required human second reviewer and logged search strings. - **Core Program Team:** no direct action; if a program relies on external research, ask for the search protocol. - Mobile: no direct action from this source. - Playground: no direct action from this source. - Community: no direct action; nearest relevance is surveys and literature scans for WordCamp research talks. - Training: build a short lesson on PICO and PRISMA basics for contributors who write research-backed proposals. - Marketing: no direct action from this source. - Support: no direct action; a scoped review of support-ticket themes could use PICO-style questions. - Design: no direct action; nearest relevance is UX research syntheses that would benefit from documented search strings. - Documentation: add a "how we searched" pattern (sources, dates, strings) to any doc that cites research. - Photos: no direct action from this source. - TV: no direct action from this source. - Polyglots: no direct action; the source is English-only and the tool's language coverage isn't stated. - Accessibility: a systematic review of accessibility-audit tool accuracy is a good pilot, with a human reviewer screening every record. - Meta: no direct action from this source. - CLI: no direct action from this source. - Hosting: no direct action from this source. - Tide: no direct action; nearest relevance is a review of published false-positive rates in code scanners. - Openverse: no direct action; nearest relevance is a review of moderation-classifier evaluations. - Themes: no direct action from this source. - Plugins: no direct action; nearest relevance is a review of plugin-security research, screened by a human. #### The WPOCC - Programs: if The WP Open Community Collective funds research fellowships, ask that funded reviews publish PRISMA counts and search strings. - Projects: no direct action from this source. - Funding experiments: a small grant for an AI-assisted evidence review could require a human second reviewer and a citation spot-check report. ### FAIR - Technical independence: a hosted review tool ties evidence to a vendor. Keep search strings and included-study lists in an exportable format. - Policies and covenants: if FAIR policy cites evidence, require the protocol (eligibility criteria and search strings) to be published with it. - Productized governance artifacts: add a "review method" row to scorecards: human reviewers count, tool used, and whether PRISMA counts are published. ## 🧪 Hypotheses 1. On a single 200-record benchmark that a human has already screened, AnswerThis title and abstract screening reaches at least 95% recall for included studies. Systematic reviews prioritize recall, and the video reports no figure. 2. The citation double-check setting (00:01:24) reduces nonexistent or misattributed citations in generated reviews to under 2% of references, compared with the same tool with the setting off. 3. Figures the tool generates from extracted data match the values in the cited papers in at least 90% of data points (for example the maximum reported PCE chart). 4. Time to a screened, human-verified systematic review falls by at least 40% versus a manual workflow, once verification time is counted. 5. Reviewers who see AI-suggested PICO criteria anchor on them: fewer than 25% of reviewers materially edit the suggestions, which would show the "critique instead of generate" workflow (00:10:47) narrows inclusion criteria. 6. The 240-to-203 duplicate removal step captures at least 98% of true duplicates when compared with a reference manager's deduplication. ## 📊 Methodology - **Design for H1 and H6.** Take a published Cochrane or PRISMA-registered review with a known included-study list. Rerun its research question through AnswerThis. Compare the tool's screened set to the published set and report recall, precision and duplicates caught. Have two human raters resolve disagreements. - **Design for H2.** Generate 10 literature reviews on 10 topics, half with citation double-check on, half off. Verify every reference against DOI and publisher pages by hand. Count missing, misattributed and unsupported claims. - **Design for H3.** Sample 10 generated charts. Trace every plotted value to the source table or paragraph in the cited PDF. Report the match rate and note unit and definition errors. - **Design for H4.** Time three review teams on the same question: manual, tool with full verification, and tool with auto-run then verification. Log active hours, not calendar days. - **Design for H5.** Show 30 reviewers a blank PICO form or an AI-drafted one. Compare edit distance between drafts and final criteria. Preregister the 25% threshold. - **Controls.** Fix the topic and database set, disable web search in one arm, and use two blinded reviewers for ground truth. Record the tool version and date, because behavior changes between releases. ## ❓ Questions - What recall and precision does the screening step achieve against a human-screened set, and on which topics? - How does the citation double-check work: DOI resolution, full-text matching, or an LLM judgment? What's its false-pass rate? - What does "PRISMA compliant" mean here (00:06:26): a flow diagram, a checklist output, or protocol registration support? - Are the generated figures computed from extracted data tables, or drawn from a language model's reading of the text? - What happens to uploaded PDFs and libraries: retention, training use, and deletion? - How blinded is the human second-reviewer mode (00:07:54), and does the tool record inter-rater agreement? - Which risk-of-bias instrument does the tool use, and is it valid outside medicine, for example on spaced-repetition education studies? - Is the Q1 and Q2 journal filter based on a named ranking such as SCImago, and how current is it? ## 🔮 Future research - **Independent screening benchmarks.** There's no public recall figure in this source. A shared benchmark of past systematic reviews would let tools be compared on the metric that matters. - **Citation hallucination rates by setting.** The double-check feature needs testing across fields with sparse coverage in Semantic Scholar. - **Automation and reviewer anchoring.** Whether AI-suggested criteria change what human reviewers include is untested for these tools. - **Risk-of-bias transfer.** Tools built for clinical trials are being run on education and materials science. Validity across those domains needs study. - **Figure provenance.** A standard for labeling machine-generated charts with the extracted data behind them would make them checkable. ## 💡 Implications - **Practice.** Use the tool for scoping and drafting, then repeat screening with a human on a sample. The narrator says this at 00:08:37 and 00:11:31. - **Policy.** Journals and funders that require PRISMA will see more flow diagrams generated by software. Requiring the search strings and reviewer logs, not just the diagram, keeps the standard meaningful. - **Security.** Uploading unpublished manuscripts or embargoed data to a hosted tool is a disclosure. Check terms first. - **Education.** The PICO-first flow teaches question framing. Students who accept the AI's criteria without editing skip the skill the exercise exists to build. - **Funding.** Review labor is often the cost driver. If a tool cuts it, grants can shift toward verification and second-reviewer time, but only if accuracy holds. ## 🧠 Learning styles assessment - **Reading level estimate.** Flesch-Kincaid Grade Level on the transcript body: about grade 8 to 9, based on short spoken sentences. Flesch Reading Ease lands near 65. Domain terms (PICO, PRISMA, Q1, risk of bias, LaTeX) raise effective difficulty for viewers outside research. #### 👀 Visual learners - Present: screen recording of every step, the generated PCE chart, the live PRISMA flow diagram and the risk-of-bias key. - Improve: no static diagram of the whole pipeline (question, PICO, strings, fetch, screen, extract, bias, PRISMA). Add one, and read chart values aloud. #### 🔊 Aural learners - Present: continuous narration with enthusiasm markers and jokes. - Improve: filler ("blah blah blah," "do a thing, go to the next one" at 00:09:20) hides the step names. Name each stage aloud in order, and offer a short audio recap. #### 🗣️ Verbal learners - Present: PICO is spelled out (population, intervention, comparator, outcome) and the workflow is described in plain speech. - Improve: PRISMA, risk of bias and Q1/Q2 go undefined. One sentence each would help newcomers. #### 🧩 Kinesthetic learners - Present: live demo of settings and a pre-run example. - Improve: no starter topic, no downloadable checklist, and the narrator skips the repetitive steps ("we won't go through them now"). Offer a scripted practice run. #### 🧠 Logical learners - Present: the systematic review order maps to the standard method, and the PRISMA counts (240, 203, 29) give a traceable funnel. - Improve: the funnel between 203 and 29 isn't shown. Show each exclusion reason and count. #### 🧑‍🤝‍🧑 Social learners - Present: comment prompts (LaTeX pronunciation), second-reviewer invite feature, and a community waitlist. - Improve: no discussion prompt about reviewer agreement or how labs split screening. #### 🧘‍♀️ Solitary learners - Present: self-contained and rewatchable, with eight chapters in the description. - Improve: add a reflection prompt, such as "write your PICO before the AI does." #### 🌿 Naturalistic learners - Present: no explicit nature content; the example topics (transparent electrodes for solar cells, asthma) touch energy and health. - Improve: link reproducible search strings to evidence-based environmental and health work. ## 📊 Accessibility assessment #### Screen reader semantics - Present: chapters are timestamped in the description and the auto-transcript is machine-readable. - Improve: chart contents and PRISMA numbers are shown visually and only partly spoken. WCAG 2.2 AA 1.1.1 (Non-text Content) and 1.2.5 (Audio Description, Prerecorded) apply to the PRISMA panel and the PCE chart. #### Keyboard navigation - Present: the YouTube player supplies keyboard controls and visible focus. - Improve: nothing in the video adds barriers; 2.1.1 (Keyboard) and 2.4.7 (Focus Visible) pass through the platform. The narrator notes the AnswerThis interface doesn't make the next click obvious at 00:09:20, which suggests a 2.4.3 (Focus Order) and 3.3.2 (Labels or Instructions) check on the product itself. #### Contrast and readability - Present: not verifiable from a transcript; screen-recorded app UI is likely light-themed. - Improve: dense settings panels risk failing 1.4.3 (Contrast Minimum) and readability at mobile size. Zoom into settings while narrating them. #### Text resizing and reflow - Present: the player scales and captions resize independently. - Improve: text baked into the recording doesn't resize or reflow, which fails 1.4.4 (Resize Text) and 1.4.10 (Reflow) for the on-screen content. #### Media and captions - Present: auto-captions exist and eight chapters help navigation. - Improve: captions garble names: "answer this" for AnswerThis, "Zotterero," "Mendlay," "archive" for arXiv, "Prisma" for PRISMA, "latex or latex." That's a 1.2.2 (Captions, Prerecorded) shortfall for proper nouns. Human-corrected captions fix it. #### Forms and interactions - Present: none in the video; the description lists three signup links. - Improve: 3.3.2 applies to the linked signup forms. Put links in the description as plain text, which the creator already does. #### Standards reference - Primary gaps against WCAG 2.2 AA: **1.2.2** (inaccurate captions for proper nouns), **1.2.5** (no audio description for on-screen charts and counts), **1.1.1** (chart data visual only), **1.4.4 / 1.4.10** (burned-in UI text). Captions are the cheapest fix. ## 🌍 DEIB assessment #### Cultural breadth - Present: examples span materials science (transparent electrodes), medicine (asthma) and education (spaced repetition). The sources listed include arXiv and open-access repositories. - Improve: no mention of non-English databases or language coverage. Systematic reviews with English-only search introduce language bias, which the video doesn't address. #### Inclusive language - Present: plain second-person address, no exclusionary terms in the transcript. - Improve: "bloody fantastic" and superlatives such as "very impressed" are idiomatic (Australian and British English) and add no facts. Replace them with the feature's measured behavior. #### Neurodiversity support - Present: eight chapters with descriptive titles, and a predictable structure of settings, results, writer, then systematic review. - Improve: filler at 00:09:20 ("blah blah blah") and the rapid "here, then here, then something here" pointing at screen elements are hard to follow. Narrate button names. #### Representation - Present: one narrator, a creator with an academic-tips audience. - Improve: no researcher outside the creator's perspective and no user with a different discipline, career stage or access need. A second reviewer voice would balance the pitch. #### Psychological safety - Present: the narrator says to check everything and cautions against auto-run for real work, which protects newer researchers. - Improve: "cannot see a better way" can make early-career researchers feel that manual review is obsolete. State that human screening is still the standard. ## ✅ Logical validity #### Strong reasoning - **Procedural argument, 00:06:26.** A process with search strings you can rerun in several databases is auditable. That's sound: reproducibility comes from visible protocol steps. Clarify by stating what the audit trail contains (strings, screening decisions, timestamps). - **Risk-aware caveat, 00:08:37.** Auto-run is fine for exploration and wrong for submission. The conclusion follows from the fact that unreviewed AI output can't carry authorship accountability. - **Design argument, 00:10:47.** Prompting the user to critique AI-drafted PICO, rather than write it from nothing, saves effort while keeping the human in control. This is plausible; the anchoring risk noted in Hypothesis 5 needs testing. - **Feature evidence, 00:10:04.** Live PRISMA counts let you check attrition by stage. The narrator shows the panel (240 found, 203 after duplicates), so the claim is demonstrated, not just asserted. #### Fallacies - **Hasty generalization, 00:12:14.** "in terms of generating a systematic literature review easily I cannot see a better way of doing it right now." One topic (spaced repetition) and one run support a claim about all reviews. *Fix:* name the tools compared and report a test on several topics. - **Unsupported superlative, 00:00:41.** "one of the best systematic and literature reviews that I've seen any tool do at the moment." No comparison set or criteria are given. *Fix:* list the tools tested and score them on stated measures. - **Appeal to the tool's own label, 00:06:26.** "apparently this is Prisma compliant." The claim rests on the vendor's label, and a flow diagram isn't the same as full compliance with the 27-item checklist. The word "apparently" partly hedges it. *Fix:* check the generated review against the PRISMA checklist item by item. - **Novelty claim without survey, 00:02:51.** "this is the first literature review that I can think of that actually creates its own figures." The narrator hedges with "that I can think of." *Fix:* soften to "the first I've tried." - **Missing evidence on accuracy, throughout.** The video shows no check that screened papers were correctly included or excluded, nor that citations exist. Praise of "impressive" output stands in for validation. *Fix:* show a sampled verification. - **Undisclosed or unstated interest.** The description sells the creator's own courses and newsletter. That doesn't prove bias, and no AnswerThis affiliate link appears in what was captured, but the video doesn't say whether the tool or its maker supplied access or payment. *Fix:* a one-line disclosure stating sponsorship status. #### Tone of voice - Enthusiastic, informal and demo-led, in a creator-tutorial register. Contractions and second person throughout, with frequent affirmations ("fantastic," "I love it"). Self-correction humor at 00:08:37 ("search strings, not streams") adds candor. The closing at 00:12:14 shifts to a strong verdict, which is where the unsupported superlatives cluster. The middle, where settings and caveats are shown, is the most careful section. ## 🛡️ DARVO analysis Not applicable to this source. DARVO (Deny, Attack, Reverse Victim and Offender) describes a response to accusation or criticism. This is a product walkthrough with no accuser and no dispute. For completeness across the three prongs: **Deny** — no denial appears; limitations are volunteered ("you shouldn't do that if you're looking to submit this anywhere," 00:08:37, and the unclear click path at 00:09:20). **Attack** — no critic or competitor is attacked. **Reverse victim/offender** — no such move appears. No related emotional or rhetorical pressure beyond enthusiastic praise. ## 🎯 Conclusion The durable idea is the workflow, not the vendor: a systematic review tool should expose editable criteria, readable search strings, staged screening and live attrition counts, so the human can audit each step. AnswerThis, as shown, has those pieces, and the narrator's caution against auto-run for anything submitted is the right rule. The evidence for "best available" is thin. It comes from one run on spaced repetition (29 studies included), one literature review on transparent electrodes, and the narrator's impression. No screening accuracy, citation-check performance or chart-to-source match is tested on camera, and the "PRISMA compliant" label is the vendor's. **Next steps for readers:** run your own question through the tool and treat the output as a draft. Screen a 50-record sample by hand and compare. Spot-check five citations and the chart values. Check the generated review against the PRISMA checklist. Read the terms on uploaded PDFs. If you cite this video, note that the creator sells their own courses and that no AnswerThis sponsorship status is stated. ## 🔄 Refactor Appendix — This AI Does Your ENTIRE Systematic Literature Review in Minutes (PRISMA Included) > Density: 2–5 Literature atoms (H3 pad), 0–3 Permanent candidates (H4 pad). > Emitted: 4 Literature, 1 Permanent. Resonance filter applied. Chapter-by-chapter feature > descriptions were dropped as product tour, not claims. > > Konik claim-header rules: present tense, declarative, ≤10 words, searchable, unique. > > Atomicity: one declarative sentence per atom. "and"/"also" → split. > > Tag pool: #bmf · #addendum · #xref · #articleseed · #storystem · #share · #aphorisms ### 📚 Literature Split Pad #### [H1] Reproducible search strings make AI review auditable > [!quote] [00:06:26] > it's not just a black box of AI doing stuff and giving you a response. This is completely auditable at the end **Context & use**: - 00:06:26, right after the tool is described as PRISMA compliant with search strings reusable across databases. - This is the design principle worth keeping. The tool's accuracy is still untested in the video, so keep the claim about process, not results. **Tags**: #xref #articleseed <!-- atomicity: 1 idea ✓ --> #### [H2] Generated figures cite the data behind them > [!quote] [00:02:51] > this is the first literature review that I can think of that actually creates its own figures. **Context & use**: - 00:02:51, before the PCE chart demo at 00:03:34. - A novelty claim with a built-in hedge ("that I can think of"). Whether chart values match the cited papers is unchecked. **Tags**: #xref <!-- atomicity: 1 idea ✓ --> #### [H3] Auto-run is for exploration, not submission > [!quote] [00:08:37] > Obviously, you shouldn't do that if you're looking to submit this anywhere. **Context & use**: - 00:08:37, about the "auto run all steps" option. - The clearest responsible-use rule in the video. Keep it beside any praise of speed. **Tags**: #xref #share <!-- atomicity: 1 idea ✓ --> #### [H4] Human second review stays part of the method > [!quote] [00:11:31] > This would have taken ages and obviously with a systematic review you go through it with someone else and you check everything **Context & use**: - 00:11:31, on the finished 29-study review. - The narrator credits speed but still assumes a human co-reviewer. The claim to test is how much of the "took ages" gap survives once verification is counted. **Tags**: #xref #articleseed <!-- atomicity: 1 idea ✓ --> ### 🧠 Atomic Split Pad #### [P1] Automate the trail, not the judgment **Claim** (one sentence, Courtney's voice): An AI-assisted review is only as trustworthy as the audit trail it leaves for a human to check. **Evidence pointer**: [[#H1]] · [[#H3]] · [[#H4]] **Scope & conditions**: Holds when search strings, screening decisions and counts are visible and editable. Breaks when the tool hides its decisions, when users click auto-run and submit, or when nobody spot-checks citations. Accuracy still needs its own test; an auditable process can still produce wrong screening. **Links**: - [[Reading vendor benchmarks]] — because contrasts on evidence: this video shows a workflow, not a measured accuracy - [[PRISMA reporting]] — because extends - [[Human in the loop review]] — because applies-in evidence synthesis <!-- atomicity: 1 idea ✓ --> ### ✏️ Fleeting Queue - [F1] Find AnswerThis's official site, pricing and terms on uploaded PDFs before citing this video — https://www.youtube.com/watch?v=CRIYKjTOJA8 #fleeting #inbox - [F2] Open question: what recall does the screening step hit against a human-screened published review (see Hypothesis 1)? #fleeting #inbox - [F3] Check whether "PRISMA compliant" means the 27-item checklist or only the flow diagram #fleeting #inbox - [F4] Article seed: "Automate the audit trail, not the judgment" for evidence reviews in WordPress research #fleeting #inbox #articleseed ### → Things 3 (max 2 actions) - [ ] Run one small question through AnswerThis and hand-verify 5 citations and 50 screened records — research, 45 min #things - [ ] Look up the PRISMA 2020 checklist and compare it with the tool's generated review — evidence review method #things <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/This AI Does Your ENTIRE Systematic Literature Review in Minutes.md|This AI Does Your ENTIRE Systematic Literature Review in Minutes by Andy Stapleton]] ## Highlights - [00:02:51] highly recommend that you put it on max and then you click go. So obviously you need to put your topic here. So OPV devices, you should probably spend a little bit more time coming up with a good prompt to you know tell it exactly what you want. But ultimately then you click go. Now, we won't wait for it because I've done it for you already with um transparent electrodes and this is what it came up with. So, um I asked it for high efficiency flexible transparent composite electrodes. And the one thing that amazed me out of the bat is that this is the first literature review that I can think of that actually creates its own figures. Check this out. as I scroll. Yeah, we've got the abstract, we've got the introduction, we've got performance and that like it [00:03:34] makes sense to me. This is all great. And if I click something, you actually get to see exactly where it got it from. And if it's got a PDF, it will actually highlight in the PDF where you got that where it got that information from to generate the literature review. And then if I keep scrolling, keep scrolling. This is what was very very impressive to me. Where is it? This. So maximum reported PCE. You can see that it's actually generated this graph and you can see it here. This is a PNG generated from this answer. This is the first time I've seen this. I'm very very impressed with it. And then you can see figures. These are the figures. You can download them and you can download a load of different things. Now, the one thing I ([View Highlight](https://read.readwise.io/read/01m3psk4y61z2h0srsxeg9as88)) - Note: [H2] Generated figures cite the data behind them - Tags: atom/literature/h2 - [00:05:43] on the promise of asking a question to submission because you can do everything that you want in here and more. And you can even export it to your favorite tool. Oh, I love it. But it gets even better because I'm about to show you probably one of the most impressive systematic literature reviews I've generated with an AI tool. That's a big claim, but this is what it looks like. Systematic reviews can take months and they should take a good amount of time so that you can actually go through the process. But the thing about using an AI tool like answer this is that the process is made incredibly streamlined and simple for you. Check this out. So if you click on systematic review, you [00:06:26] immediately get taken to another interface. And here you can see you can start a new run. So create a reusable systematic review workspace. So apparently this is Prisma compliant which means that it will give you all of the search strings that is search and you can do that reproducibly in a number of databases so it's not just a black box of AI doing stuff and giving you a response. This is completely auditable at the end which I really love. So here you save it to a default project. Yes. And then put your research questions here. So, let's just choose one of their research questions and then go create review. And you'll see that you get taken to this area where you go through all of the steps that you would do ([View Highlight](https://read.readwise.io/read/01m3psk2hvvt09kccphkhf42cj)) - Note: [H1] Reproducible search strings make AI review auditable - Tags: atom/literature/h1 - [00:08:37] so you can see that uh we can regenerate them with AI or we can go in and actually just generate some ourselves we can change it fantastic you can skip all of these steps if you just are like you know what I just want an output just to see what it does. You can go auto run all steps. Obviously, you shouldn't do that if you're looking to submit this anywhere. If you're going to use it for proper research, but I love the fact that there is just like a do it for me and let's see what the outcome is. I think that's super important if you just want to test something. You want to see what this tool does, what it's like. And uh yeah, I I I think it's fantastic. So, here here's the search streams uh that it's come up with. And search strings, not streams. Yeah, I caught myself. [00:09:20] That's good, isn't it? Live editing. And then we've got here fetch papers. So, you go through all of the steps one at a time. These are all of the steps. We won't go through them now because they do get, you know, a little bit boring and repetitive, but ultimately here, fetch papers, go to the next one, do a thing, go to the next one. It's not always super obvious where you need to click to actually get it to do stuff, but here, then here, then something here, button here, and then here. And you can see we go through full text screening, title and abstract screening, data extraction, risk of bias, blah blah blah. It just goes on and on and it guides you through the process. So this is what I was able to generate. Um, and actually while I before I go there, let's go back. Before I get there, the ([View Highlight](https://read.readwise.io/read/01m3psk7b2df08fk2wy17ankc0)) - Note: [H3] Auto-run is for exploration, not submission - Tags: atom/literature/h3 - [00:11:31] I go to workspace, you can see this is everything I went through. All of these were ticked and uh the Prisma diagram I can look at which is fantastic. But if I want to go to the very end and the actual systematic literature review, this is where I can go and you can see we've got abstract background methods. But this is where it starts to get very very impressive because here's the Prisma flow diagram completely made for me. It was 29 studies including this and then we've got the risk of bias and including the sort of method that we used to identify bias and it's even got a key down here. This would have taken ages and obviously with a systematic review you go through it with someone else and you check everything but in terms of generating a systematic [00:12:14] literature review easily I cannot see a better way of doing it right now. And look these are all of the studies that have been included. It's got lots of writing. It's got all of the data. It's got all of the things. It's got a discussion, summary of findings, strengths and limitations of this review, comparison with prior reviews. It's just got everything you would want to make a start on your own sort of like uh systematic literature review. It gives you the ability to generate one, start arguing with it. And the fact that it's so customizable as you go along just means I think it is just a fantastic way to work if you have to generate a systematic literature review. Well done. Answer this. ([View Highlight](https://read.readwise.io/read/01m3pska1q6gj4naefnnpezyet)) - Note: [H4] Human second review stays part of the method - Tags: atom/literature/h3