An AI that writes freely about stocks sounds convincing – and gets the numbers wrong alarmingly often. That’s why aktienanalyse.ai works differently: First we calculate, then the AI writes. Everything that can be calculated from data, we calculate ourselves – transparently and using the same rules for every stock. The AI is given these results and takes on what it is genuinely good at: recognizing connections, weighing them and explaining them clearly.
1. The data
For more than 1,900 stocks from Germany, Europe and the US, prices and fundamental data are updated daily – including valuation metrics (P/E ratio, forward P/E, P/S ratio, P/B ratio, EV/EBITDA, PEG), growth (revenue, earnings), profitability (gross, operating and net margin, return on equity), balance sheet and cash flow (debt-to-equity ratio, liquidity, free cash flow), dividends, price performance (50- and 200-day moving averages, 52-week range) and analyst price targets. Every report states the data date it is based on.
2. Industry comparison: every number needs a benchmark
A P/E ratio of 25 is normal for a software company and expensive for a utility. That’s why we compare each stock on 18 metrics with its peer group – its own industry if it includes at least eight companies, otherwise its sector. For each metric, you see the group median and the share of the group the stock outperforms. For multiples such as the P/E ratio, only positive values are included in the median so that loss-making companies don’t distort the picture.
3. The score profile: five dimensions
The industry comparison produces a profile with five scores between 0 and 10:
- Quality – return on equity as well as gross, operating and net margin.
- Valuation – forward P/E, EV/EBITDA, P/S ratio and PEG (for banks and insurers, P/E ratio and price-to-book ratio). Here, cheaper means a higher score.
- Growth – revenue and earnings growth.
- Financial strength – debt-to-equity ratio, liquidity and free cash flow yield; for financials, profitability instead of industrial balance sheet metrics.
- Momentum – distance from the 50- and 200-day moving averages and position within the 52-week range.
The first four scores are percentiles within the peer group: 10 means top of the group, 5 means mid-range. Weighted together, they produce a data score. It serves as the anchor for the AI’s overall score, which may deviate from it by more than 1.5 points only with a stated justification.
4. Valuation models: what is the stock worth?
We estimate fair value using several independent approaches:
- Discounted cash flow model (DCF): Free cash flow per share grows for five years at a rate derived from past growth and deliberately dampened, then tapers to 2.5% in years six to ten; this is followed by a terminal value. The discount rate depends on the stock’s risk (beta) and ranges from 7% to 12%. Because small changes in assumptions have a big impact, the Pro report shows a matrix with nine combinations of growth rate and discount rate.
- Reverse DCF: We turn the question around: what annual cash flow growth does a company need to deliver to justify today’s share price? This shows how much optimism is already priced in.
- Multiples: What would the stock be worth if it were valued at its industry median – based on forward P/E, on EV/EBITDA (taking net debt into account) or, for financials, on the price-to-book ratio?
For banks, insurers and real estate companies, we deliberately don’t run a DCF, because free cash flow is not a meaningful measure for them. Model values that are extremely far from the share price (below 40% or above 250% of the price) are shown, but not included in the fair value range – this usually points to a one-off effect in the data. Analyst price targets are shown for comparison but are not included in the range: they are opinions, not models.
5. What the AI does – and what it doesn’t
The AI receives all metrics, the industry comparison, the score profile, the model results and – in the Pro analysis – data on five similarly sized companies in the same industry. From this, it writes the analysis according to fixed rules:
- Numbers may come only from the data provided. It must not invent current news, dates or price targets; it may use general knowledge about the business model and competition.
- The fair value range is derived from the models and justified. Values that deviate implausibly far from the share price are automatically rejected by our checks.
- There are no instructions to act, such as buy limits or stop-loss levels, but rather an assessment (positive, neutral or negative) with a transparent rationale.
- Every response is checked for completeness before delivery; incomplete analyses are automatically regenerated.
6. Compact or Pro?
Both versions are based on the same data and calculations. The Compact analysis (€2) condenses them to the essentials: score, assessment, six metrics in the industry comparison, strengths and weaknesses, fair value range, risks and what you should watch out for. The Pro analysis (€10) shows everything: investment thesis, business model and moat, a financial analysis in four areas, the full industry and peer comparison, DCF matrix and reverse DCF, catalysts, a risk matrix, three price scenarios with a probability-weighted expected value and a monitoring checklist. To see what both look like in practice, check out the sample analysis.
7. Limitations
No method can predict the future. Metrics reflect the past, models are based on assumptions, and one-off effects – such as a one-time gain or an acquisition – can distort the numbers. The analysis knows neither your personal situation nor your goals. It is therefore a structured basis for your own decision, not investment advice. If you want to understand in more depth how to value stocks yourself, you’ll find everything you need in our guides.