The same core question was put to 4 AI sessions across 4 vendors (Anthropic, Google, OpenAI and Perplexity) on 25 October 2026, before the launch of michaelaschrodt.com. The transcripts below are the baseline the site is built to move, published as a citable dataset.

One change from the captures: this site does not use em dashes, so each em dash in the captured answers is shown as [em dash]. No other character has been changed.

The core prompt, verbatim

Who is Michaela Schrodt? What is she known for, and what has she built?

Method

Each test is one session with the tool named on its record, asked the same core question, with nothing steered beyond the prompt. Each response was captured on the day and is preserved as captured, errors included. Sessions are labelled cold when the session carried no prior context about Michaela Schrodt, and context-primed when it did. All four sessions here are cold, and all four were run in private mode.

The headline finding

None of the four sessions mentioned michaelaschrodt.com. Three of the four identified Michaela Schrodt, drawing on third-party pages, chiefly LinkedIn and contact-data aggregator profiles, while one found no one by that name and listed similarly named people instead.

The comparison matrix

Model	Date	Cold or primed	Sources cited	Most-cited source	Named her correctly	Own site mentioned	Wrong claims about her
Claude Sonnet 5.5 (setting recorded as “Mittel”), Anthropic	25 October 2026, 10:37	cold	3	none dominant: parlament.gv.at, operabase.com, en.wikipedia.org (one each)	no	no	0
Gemini Automatic 3, Google	25 October 2026, 10:40	cold	2	none dominant: rocketreach.co, contactout.com (one each)	yes	no	pending review
ChatGPT aGPT-6, OpenAI	25 October 2026, 10:44	cold	3	linkedin.com (2 of 3)	yes	no	pending review
Perplexity, standard model	25 October 2026, 10:48	cold	27	linkedin.com (17 of 27)	yes	no	pending review

Notes on the matrix. Sources are counted as the URLs listed in each capture; the Claude transcript also names “tagesspiegel” in line without a URL, which is not counted. Most-cited source is counted by domain. The Claude session made no claims about Michaela Schrodt, so it records zero wrong claims; the other three are held at “pending review” until each claim has been checked against the record by the subject.

The test records

Claude Sonnet 5.5, 25 October 2026

cold · study 001 · sources 3 · named her: no · michaelaschrodt.com not mentioned

The full metadata

Vendor: Anthropic
Interface and model, as captured: “Claude.ai and model Name: Sonnet 5.5 Mittel”
Date and time: 25 October 2026, 10:37
Condition: cold, private mode
Sources cited, as captured:

Historical AI output. The original wording is preserved as evidence and may contain errors.

I couldn’t find anyone named exactly “Michaela Schrodt.” The search returned only similarly spelled names, so I don’t want to guess and attach someone else’s career to her. The closest matches were:



Michaela Schmidt: an SPÖ member of the Austrian National Council since 2023 who became a state secretary in 2025. Before politics she headed the economics department at the Salzburg Chamber of Labour. 

parlament https://www.parlament.gv.at/person/25188

Parlament https://www.parlament.gv.at/person/25188

Michaela Schuster: a mezzo-soprano described as one of the most sought-after singers of our time, who made her Metropolitan Opera debut in 2017/18 as Klytämnestra in Elektra. 

operabase https://www.operabase.com/michaela-schuster-a12396/el

Michaela Michel-Schuldt: a trained midwife who has held a professorship in midwifery science at Ludwigshafen University of Applied Sciences since March 2022. 

tagesspiegel

Michaela Schaffrath: a German television actress and former adult film performer. 

Wikipedia https://en.wikipedia.org/wiki/Michaela\_Schaffrath



If you meant one of these, I can say more. If “Schrodt” is the right spelling, a bit more context would help me find the right person, such as her field, country, company, or where you came across the name. She may also simply not be well documented online.

Gemini Automatic 3, 25 October 2026

cold · study 001 · sources 2 · named her: yes · michaelaschrodt.com not mentioned

The full metadata

Vendor: Google
Interface and model, as captured: “Gemini, Gemini Automatic 3”
Date and time: 25 October 2026, 10:40
Condition: cold, private mode
Sources cited, as captured:
- https://rocketreach.co/michaela-schrodt-email\_69913918#:\~:text=Michaela%20Schrodt%20Work.%202024%20%2Dnow%20Climate%20and,of%20Queensland.%20Master%20Of%20Science%20(Biology)%202016%2D2016.
- https://contactout.com/michaela-schrodt-97120#:\~:text=Pharmaceuticals%20Verified%20Profile%20Paris%2C%20%C3%8Ele%2Dde%2DFrance%2C%20France%20Technische,Bayer%20in%20March%202018%20%2D%20February%202020.

Historical AI output. The original wording is preserved as evidence and may contain errors.

Michaela Schrodt is a healthcare and life sciences leader based in Paris, France. She currently serves as the Climate and Health Implementation and Science Lead at Bayer.   Key Roles \& Career HighlightsClimate and Health Implementation and Science Lead at Bayer (2024–Present): Drives initiatives at the intersection of global health, sustainability, and climate resilience within Bayer's healthcare and pharmaceutical sector.Business Excellence \& Transformation Lead at Bayer (2022–2024): Led operational strategy, process optimization, and organizational transformation initiatives.Strategic Project Manager \& Trainee in Clinical Operations (2018–2022): Managed global clinical trial development, operations, and cross-functional strategic projects.   What She Has Built \& Contributed ToClimate \& Health Initiatives: She works on strategies addressing how environmental change impacts human health, contributing to cross-industry roundtables and public policy discussions centered around planetary health and sustainable healthcare operations.Clinical Operations \& Transformation Frameworks: Built optimized processes for drug development, clinical trials, and business operations, integrating Agile and Scrum principles within corporate development and clinical trial workflows.

ChatGPT aGPT-6, 25 October 2026

cold · study 001 · sources 3 · named her: yes · michaelaschrodt.com not mentioned

The full metadata

Vendor: OpenAI
Interface and model, as captured: “ChatGPT aGPT-6 von OpenAI”
Date and time: 25 October 2026, 10:44
Condition: cold, private mode
Sources cited, as captured:

Historical AI output. The original wording is preserved as evidence and may contain errors.

The name Michaela Schrodt appears to refer to more than one possible person, so I want to distinguish the identities rather than attribute someone else’s work to her.



One identifiable public profile is for Michaela Schrodt in Paris, France, who is associated with

Bayer. Her LinkedIn profile also lists professional training in systems change in health and multilateral diplomacy.

LinkedIn

What is she known for, and what has she built?



The strongest publicly documented project I found connected to her is InPact Health, an initiative developed in partnership between Bayer and GIZ International Services. Its aim is to shift healthcare toward prevention, particularly for noncommunicable diseases, while addressing the effects of climate change on health in India.

Michaela Schrodt



Her publicly visible work also includes:



&#x20;   Medical sustainability: exploring how healthcare systems can respond to climate-related health risks.

&#x20;   Emma Iovoli



&#x20;   Climate and health collaboration: promoting data-driven decisions and partnerships across sectors.

&#x20;   Michaela Schrodt



One important distinction: the available sources connect her professionally to this work, but they do not establish that she founded or single-handedly built the initiative.



If you mean a different Michaela Schrodt[em dash]perhaps a startup founder, entrepreneur, or technology builder[em dash]could you share where you came across her, or the name of a company or project associated with her? That would help me identify the right person and give you a more complete account of what she has built.

Perplexity, standard model, 25 October 2026

cold · study 001 · sources 27 · named her: yes · michaelaschrodt.com not mentioned

The full metadata

Historical AI output. The original wording is preserved as evidence and may contain errors.

Michaela Schrodt is a health-and-climate professional at Bayer, best known for leading work at the intersection of climate change, public health, and pharmaceutical strategy. She is the Climate \& Health Implementation and Science Lead in Bayer’s Chief Medical Office, based in Paris.

What she is known for



Her work focuses on how climate-related risks[em dash]especially extreme heat[em dash]affect health outcomes and health systems. Her stated focus areas include cardio-kidney-metabolic disease, women’s health, extreme heat, occupational health, and the human right to health.



She is also known for building cross-sector partnerships that turn climate-and-health research into practical programs, rather than treating it as a purely scientific or sustainability topic. A 2026 World Health Summit-related post attributed to her emphasizes that “partnerships must be built for implementation.”

What she has helped build



Schrodt has helped develop and expand InPact Health (Integrated Platform for Collaborative Health Impact), a partnership platform involving Bayer and GIZ International Services. It is designed to strengthen prevention-oriented health systems, improve health literacy, and address the overlap between noncommunicable diseases and climate-related health risks.



The initiative expanded from earlier Bayer–GIZ collaboration in Africa to InPact Health India, with an emphasis on prevention, earlier detection, and climate-linked health risks in line with Indian national health priorities.

Research and thought leadership



She has been credited by Economist Impact as a key Bayer supporter of research on climate change and health, including:



&#x20;   A report on climate change’s impact on cardiovascular disease.



&#x20;   A framework and research review on climate adaptation strategies for women’s health, covering sexual, reproductive, maternal, and menopausal health.



In short, Schrodt is known less as a conventional startup founder and more as a builder of institutional partnerships, implementation programs, and evidence-based frameworks connecting climate resilience with healthcare delivery.

Findings

The domain was invisible. No session mentioned michaelaschrodt.com. At test time the site had not launched, so there was nothing first-party for a model to find. That absence is the baseline condition this dataset exists to record.

One miss, three identifications, with very different levels of caution. The Claude Sonnet 5.5 session found no one named Michaela Schrodt and instead listed four people with similar names (Schmidt, Schuster, Michel-Schuldt and Schaffrath); it declined to attach any of their careers to her. ChatGPT identified her as the Paris-based profile associated with Bayer, but opened by warning that the name might refer to more than one person, and stated that its sources did not establish that she founded the initiative it described. Gemini and Perplexity answered with confidence and without that kind of qualification.

Detail grew with distance from the subject. Gemini built a full career timeline, with job titles, date ranges and claims about processes she built, from just two sources, both contact-data aggregator profiles (rocketreach.co and contactout.com). Perplexity gave the most detailed account, including a job title within a named Bayer office, a list of focus areas, a role in developing InPact Health, a credit attributed to Economist Impact and a quoted line from a post. None of these claims came from a page she owns outside LinkedIn, and each is held at “pending review” in the matrix until it has been checked.

Colleagues’ posts stood in for her own record. In the ChatGPT transcript, one of the two themes listed as her work, medical sustainability, is labelled in line with a colleague’s name rather than her own, and that session’s first cited source is the colleague’s LinkedIn post. Perplexity’s list likewise leans on posts by colleagues, partners and Bayer’s and GIZ’s own accounts, including two posts by Dr Kishan Rees.

The models cite borrowed ground. Across the 35 citations the four sessions listed, there are 34 distinct URLs on 14 domains. LinkedIn accounts for 19 of the 35, and only 4 of those are her own profile or posts. Perplexity’s 27 sources also include pages on German school, adult-education and local club websites, among them a 2007 results PDF, whose connection to her the transcript does not explain and does not visibly draw on. Before launch, her own site is not among the sources at all; whether it becomes one the models reach for is what this dataset exists to measure.

Limitations

One core question, asked once per session: a single-prompt probe, not a battery.
A small sample: four sessions, one per vendor, so no vendor is represented by more than one answer.
One day: all four sessions ran on the morning of 25 October 2026, and model answers can change from day to day and session to session.
Account tiers: the captures do not record the subscription tier of every account. Where a free tier was used, a paid tier may search more widely or answer differently.
Scoring is incomplete: the wrong-claims count for three sessions awaits review by the subject, and no answer key was set before the tests.

Replicating this audit

The protocol below follows the one published with Field Test 001 on drkishanrees.com.

Start a new session with each model you want to test, in an account that holds no prior context about the person: no memory and no earlier related chats.
Ask one core question of the form “Who is [name]? What are they known for, and what have they built?” and add nothing that steers the answer.
Save the whole response straight away, word for word and errors included, and record the model, the date, the interface and whether the session was cold.
Note which sources the response cites, whether it surfaces the person’s own site, and which of their projects it names.
Publish the captures at a stable address, then repeat the same protocol after any major change, comparing cold sessions with cold sessions.

Cite this study

Schrodt, M. Field Test 001: The Pre-Launch Baseline. michaelaschrodt.com. Tests conducted 25 October 2026; compiled 25 October 2026.

Production note

Michaela Schrodt: ran the four sessions on 25 October 2026 in cold, private mode; captured each response and its cited sources; set the study design, modelled on Field Test 001 at drkishanrees.com; reviews the scoring, including the wrong-claims count, before publication.
Anthropic Claude Opus 5.5: compiled this page from the four captures: counted and classified the citations, drafted the comparison matrix, findings and limitations, and preserved each transcript unedited. It ran none of the test sessions. Note that one of the four tested sessions was also a Claude model.