AI Investor Perception Report
Put the right answer in front of investors before they ever ask you.
Every quarter we put the questions an investor actually asks to ChatGPT, Gemini, Perplexity and Claude, and record what comes back about your business, your management, your funding and your peers. You get the answers word for word, the sources behind them, every claim that is wrong, and the fixes, in time to change the answer before the next raise.
Quarterly Senior-led Five engines Investor Discovery Audit included
- Written for
- The CFO, the IR director and the board Board-grade throughout, and it ends with one slide for the pack rather than an appendix nobody opens.
- Run against
- ChatGPT, Gemini, Perplexity, Claude, AI Overviews Each queried separately, because they read different sources and reach different conclusions.
- Held constant
- The same question set, quarter after quarter A set that moves between quarters cannot show you movement. Yours is agreed once and then held.
Three stages, run once a quarter. Nothing is queried until you have signed off the question set.
You are in itAt stage one, and again on the call at stage three. Everything between is ours.
From question set to board pack, in three stages.
The work is not collecting answers. It is deciding which of the answers will cost you money, and putting them in an order somebody can act on.
-
We agree the questions
We start from the questions your investors already put to you, and add the ones they put to an engine when your name is not in the room: the sector screen, the geography, the stage. You see the full set and change anything you want changed before we run a single query. Once it is agreed, the set is held constant, because a question set that moves between quarters cannot show you movement.
-
We run it across the engines
Each question goes to ChatGPT, Gemini, Perplexity, Claude and Google's AI Overviews separately, because they read different sources and reach different conclusions. Every answer is captured word for word and dated, with whatever the engine cited recorded beside it. Then a senior practitioner reads the set. Collecting answers is automatic. Deciding which of them will cost you money is the work.
-
You get the report and the fixes
A board-grade document: the score, the answers in full, the errors with the source behind each one, and a fix list in the order the work should happen, with an owner on every line. Then we walk it through on a call, with your IR adviser on it if that helps. The next quarter runs the same questions, so the report shows you movement rather than a fresh opinion.
Six parts, in this order. Part one is the page that goes in the board pack.
The scoreThe exact average of the six dimensions printed under it, so it cannot flatter you.
The deliverable
What the engines say, what they got wrong, and what changes it.
One score for how AI engines currently present you to an investor, and underneath it the score for each part of the story with the answers that produced it. Then every finding becomes an action, ordered by what it changes against what it costs to do.
The score is not the point. Two companies can score the same and need entirely different work: one is described accurately but never surfaces on a screen, the other surfaces everywhere with a funding figure three years out of date. What you act on is the spread.
-
Part 1The perception score and the spread
One number for how the engines currently present you to an investor, out of 100, built as the exact average of the dimensions beneath it. Beside it, the spread: which parts of the story land and which do not. This is the page that goes in the board pack, and it reads in one glance.
-
Part 2The answers, verbatim
What each engine actually said, quoted in full, with the date and the engine against every answer. Nothing is paraphrased into a friendlier version. If an engine describes your business in a way you would never use, the report shows you the sentence.
-
Part 3The long-list: Investor Discovery Audit
The screening questions run without your name in them, and who the engines put forward in answer. Where you appear, your position. Where you do not, the companies that took the slot and the material the engines drew on to choose them.
-
Part 4Where the engines are wrong
Every claim that is factually wrong, out of date or misattributed, listed with the engine that made it and the source it appears to come from. Errors are separated from unflattering but accurate findings, because those two things need entirely different responses.
-
Part 5The sources behind the answers
The material the engines are reading about you: your own pages, filings, data aggregator profiles, coverage, third-party databases. Ranked by how often it turns up in answers, so you can see which handful of sources is doing most of the talking.
-
Part 6The fix list and the board slide
Every finding restated as an action, ordered by what it changes against what it costs to do, with an owner on each line: your team, your developer, your IR adviser or us. Plus one slide that states the position and the direction of travel, for the board pack.
Eight classes. The line in gold is the query itself, put to each engine as written.
Scored separatelyEach class carries its own score, so a weak answer cannot hide inside an average.
An investor does not ask one question. They ask a sequence.
The answer to each question frames the next one, which is why the set runs from the opening summary all the way to the screen that never names you. Your own set is built from these eight classes and agreed before anything runs.
-
What does yourfirm.com do?
The opening summary
The first thing any engine produces is a summary, and it is the paragraph every later answer is built on. We record what each engine says you do, who it says you serve and where it places you, then check that against your own materials. A summary that describes the company you were two funding rounds ago quietly misprices everything an investor reads afterwards.
Left aloneAn investor arrives holding a description of a company you no longer are.
-
Why would anyone invest in yourfirm.com, and why would they not?
The equity story
We ask the engines to make the case for investing, and to make the case against. What comes back shows which parts of your story have travelled and which never left your own deck. Investors form a view from this before they open a data room, and a story the engines cannot retell is a story that is not yet in circulation.
Left aloneThe bear case is fluent and the bull case is thin, in both directions.
-
Who runs yourfirm.com and what have they done before?
Management and track record
Named executives, prior roles, prior outcomes. Engines are confident about people and frequently wrong: a director who left last year still listed, a prior company misattributed, a biography merged with a namesake. We check every named individual against the record, because an investor forms a judgement about a management team long before the first call.
Left aloneA judgement is formed about your team from somebody else's career.
-
How much has yourfirm.com raised, from whom, and when?
Financial position and funding
Raises, rounds, backers, revenue claims, valuation talk. This is where stale third-party profiles do the most damage, because an old figure repeated by a data aggregator is treated as current by every engine that reads it. We record what each engine states, and trace the number back to the source that is still publishing it.
Left aloneYou negotiate against a valuation anchor you did not set and cannot see.
-
Are there any concerns about yourfirm.com an investor should know?
Risk, governance and controversy
We ask the questions a cautious investor asks in private: litigation, regulatory exposure, jurisdictional risk, anything contested. The finding that matters is not only what is said but how it is framed and how old it is. A resolved matter still described in the present tense is a live problem in an answer, however settled it is in fact.
Left aloneA closed matter reads as an open one, and nobody raises it with you.
-
How does yourfirm.com compare with its competitors?
Peers and comparison
Engines answer comparison questions readily, and the comparison sets they choose are rarely the ones a company would choose for itself. We record which companies you are put beside, which you are ranked behind and on what basis, because the peer set an engine picks becomes the peer set an investor benchmarks you against.
Left aloneYou are benchmarked against a peer set you would never have chosen.
-
Which companies lead this sector in this market?
The screen that never names you
The Investor Discovery Audit. We run the screening questions an investor asks before any company is in mind: the sector, the geography, the stage, the thesis. Then we record who the engines put forward and whether you are among them. Accurate description is worth little if the screen never surfaces you in the first place.
Left aloneThe long-list is drawn up and you were never on it.
-
Where did that come from?
Sources and provenance
Every answer is traced to what the engine cited or, where nothing is cited, to the sources that most plausibly produced it. This is what turns the report from an observation into a plan: you cannot argue with an engine, but you can correct the profile, the filing, the page or the article it is reading.
Left aloneWithout this the report is an observation rather than a task list.
Your factsheet beside the one an engine is handing an investor, line by line.
Why it worksCorrecting the source changes the engine's answer and what the analyst reading it finds.
Same company, two records. Only one of them is yours.
An AI answer about your company is assembled from material a crawler could fetch and trust: your pages, your filings, your coverage, and the third-party profiles that quietly republish all three. Where those disagree with your own record, the engine has no way of knowing which version you would stand behind.
This is why the report traces every finding to a source rather than stopping at the answer. You cannot argue with an engine and you cannot edit its output. You can correct the profile carrying the old round, publish the leadership page nobody ever wrote, or get the company into the comparison sources the engines keep quoting. Each of those is an ordinary piece of communications work, and each changes what the next investor is told.
Illustrative. Yourfirm.com is not a client.
The cheapest version of this problem is the one you find a quarter early. The expensive version is the one an investor finds first.
Why now
A wrong answer hardens.
Engines learn from what is already published, and what is already published increasingly includes what other engines have said. A figure that enters the record wrong gets repeated, quoted and summarised until it is simply what the internet says about you.
Nothing about that process is malicious and nobody is monitoring it. A funding round nobody updated, a director who left, a description written when you were a different company: each is a small clerical fact, and each is fixable at the source an investor is reading.
The moment to find out is before the raise, not during it. Corrections take weeks to work through the sources engines read, and the quarter you want the answer right is the quarter you cannot afford to start.
- A data aggregator profilecarries the round before last, never updated
- A sector round-upquotes the aggregator, because it is easy to cite
- An AI answercites the round-up, and states the figure as current
- The next answerreads the first one, and the figure is now simply the record
By the fourth step nobody involved has made a mistake, and the figure is wrong.
Still weighing it up? Ask us anything.
Questions
Questions we are asked before the first quarter.
What is an AI Investor Perception Report?
It is a quarterly record of what AI engines tell an investor about your company. We agree the questions an investor actually asks, put them to ChatGPT, Gemini, Perplexity and Claude, and capture the answers word for word alongside the sources each engine drew on. The report scores what came back, marks every claim that is wrong or out of date, and ends with a fix list and a single slide your board can read in a minute.
Do investors really research companies with AI?
Enough of them that it is worth knowing what the answer says. We do not publish a percentage for this and would not trust one that was put to us, because nobody can audit how an analyst opens their research. What we can tell you is what the engines say about you today, which is the part you can act on. If an investor never asks, you have lost nothing. If one does, the answer they get is the frame they bring to the meeting.
Why does it matter what an AI engine says about us?
Because it is increasingly the first thing an investor reads. Analysts, family offices and fund researchers now open with an engine rather than a search box, and the answer they get is a summary written from whatever the engine can find. That answer sets the frame for the meeting before anyone from your team is in the room, and unlike a search results page it shows no competing view alongside it.
What is an AI long-list?
It is the set of companies an engine puts forward when an investor screens a market without naming anyone: the sector, the geography, the stage, the thesis. The engine returns a handful of names, and those names are the long-list that investor starts from. It is drawn from whatever comparison material the engine can find, which is usually sector round-ups, industry databases and coverage rather than any company's own site.
What is the Investor Discovery Audit, and is it included?
It is included in every report. The perception half asks what engines say when someone names you. The discovery half asks whether you come up at all when nobody does: when an investor screens your sector, your geography or your stage without a company in mind. Being described accurately is no use if you never make the long-list, so the report measures both.
Which engines do you test?
ChatGPT, Gemini, Perplexity, Claude and Google's AI Overviews. Each is queried separately, because they read different sources and reach different conclusions, and the differences between them are often the most useful finding in the report. Where an engine is unavailable or rate-limited during a run, the report says so rather than quietly leaving it out.
How do you decide which questions to ask?
We start from the questions your investors already ask you: the equity story, the management record, the funding history, the risks, the peer set. We add the screening questions an investor would ask before your name comes up at all. You see the full set and can change it before we run anything, and the set is held constant between quarters so the movement means something.
What happens if an engine says something that is wrong?
It goes in the report as an error, with the answer quoted, the engine named and the source the engine appears to have drawn it from. Most wrong answers trace back to something real: a stale profile on a data aggregator, an old funding round nobody corrected, a press mention that never got updated. The fix list names the source, not just the symptom, because correcting the source is what changes the answer.
How quickly can the answer change?
It depends entirely on what is producing it. A wrong figure carried by one stale third-party profile can drop out of answers within weeks of that profile being corrected. A perception built on years of thin coverage takes a programme of work, not a correction. The report is explicit about which of your findings is which, so nobody is promised a timeline the mechanism cannot deliver.
Is this the same as the AI Visibility Dashboard?
No. The Dashboard is a live tracker for your whole market, month by month. This report is a board document written for one audience, investors, once a quarter, in the language of a raise or a listing. Clients preparing for a specific event usually take the report; clients managing an ongoing position usually take both, and the report cites the Dashboard's evidence archive when they do.
Request the report
Change the answer before the next raise.
Tell us who you are and where to reach you. We come back with the question set for your approval, and the report follows with every answer, every error and every fix.
