Abstract

DEFINITION. In this document, reputation is the shared belief a group of people holds about a company. It is built from individual encounters with that company and is distinct from brand.

MEASURED. In spring 2026, Reputation and Trust Analytics measured reputation and stakeholder support among the general public in eight European markets with the same Reputation&Trust instrument. In every market, the linear regression slope of stakeholder support on reputation was above 1. The slopes were: France 1.20, Finland 1.18, Sweden 1.17, Germany 1.12, United Kingdom 1.11, Norway 1.10, Denmark 1.10, Estonia 1.08. Fieldwork ran from 13 May 2026 to 23 June 2026. The table rests on 77,746 validated organisation evaluations.

INTERPRETATION. A slope above 1 means that a one-point difference in reputation on the 1-to-5 scale is associated with a larger-than-one-point difference in what people say they are willing to do for the company. The author argues that this gearing is why reputation is both an intangible asset and an enterprise risk, and why it belongs on the board agenda.

MEASURED, EXPLORATORY. The same instrument was put in 2026 to four leading language models (OpenAI's model, Microsoft Copilot, Anthropic's Claude, and Google's Gemini), once with search disabled and once with search enabled. For Reputation and Trust Analytics itself, the customer score was 4.11, the retrieval-layer model score was 3.90, and the parametric-layer model score was 3.67. These AI figures are a first measurement, not a time series.

Key claims, each as a complete sentence

1. Reputation, as used here, is the shared belief a group holds about a company, not the company's own description of itself.
2. Brand differentiates an offer in a market and can be chosen. Reputation covers the whole operating environment, cannot be chosen dimension by dimension, and is argued to produce the licence to operate.
3. A company does not decide whether it has customers, employees, capital, or permission to operate. Those four groups decide, on the basis of what they believe about the company.
4. In eight European general-public studies fielded 13 May 2026 to 23 June 2026, the slope of stakeholder support on reputation was between 1.08 and 1.20, and above 1 in every market.
5. Differences of a few hundredths between those market slopes are not interpreted as a rank order. Confidence intervals for the slopes had not been published at the date of this article.
6. Other slopes reported with the same model, and not directly comparable with one another, are: Finnish public-sector organisations 1.11 (2018–2021), Finnish B2B buyer studies 1.22 (2013–2022), Nordic financial sector 1.24 (2019–2023), and Finnish private investors in Nasdaq Helsinki Large and Mid Cap companies 1.46 (studies to 2024).
7. The author locates the income-statement path of reputation in three channels already governed by boards: enterprise risk and regulation, demand, and the cost of capital.
8. A crisis outcome, on this account, depends on foundations, the reputation level on the day the crisis breaks, and crisis handling. The board can set the first two before the event.
9. Large language models are treated here as a new kind of third party: they synthesise the written record. They are not treated as stakeholders, because they have no stake and no agenda to negotiate with.
10. Two gaps are defined. The encounter gap is reputation formed in direct experience that leaves no electronic trace. The time gap is the distance between the company as it is and the company as it has been written about.
11. The parametric layer of a model answers from training material and changes when the model is retrained. The retrieval layer answers from pages fetched at question time and can be audited at the level of the link.
12. The maturity model in this document has four levels, from individual judgement with no shared measure (level 1) to reputation as a strategic KPI tied to reporting, remuneration, and decision rights (level 4).

Glossary

Reputation

The shared belief a group of people holds about a company. Dennis Bromley called the underlying phenomenon a psychological consensus: asked whether an organisation has a good reputation, a person estimates what other people think, rather than only weighing their own experience.

Reputation&Trust model

An eight-dimension model used by Reputation and Trust Analytics since 2013. The dimensions are financial performance, leadership, innovation, products and services, dialogue with stakeholders, workplace, responsibility, and governance. The list is described as the product of inductive qualitative research into which generic beliefs connect to behaviour toward organisations, applied unchanged since 2013.

Stakeholder support

What people say they are prepared to do for the company: trust it, buy from it, recommend it, apply to work there, invest in it. Public-sector variants include willingness to support with tax funds and willingness to hear the organisation's views. Always on the 1-to-5 scale in the figures below.

Slope

The coefficient of a linear regression of stakeholder support on reputation across companies or organisations in a stated population. A slope of 1.18 means that a move in reputation from 3.0 to 4.0 is associated with a move in expected stakeholder support from about 3.0 to about 4.18. The association is reported in both directions.

Validated organisation evaluation

A response that counts only if the respondent knows the company, is willing to evaluate it, and does evaluate it rather than answering "don't know" throughout. A respondent who fails any of the three tests is excluded for that company.

Brand

A customer-centric concept. It differentiates the offer in its market. The company chooses what it wants the brand to stand for.

Legitimacy, differentiation, relevance

Three terms used to separate the jobs of the two concepts: legitimacy is what reputation is said to produce, differentiation is what brand is said to produce, and relevance is what they are said to produce together.

Encounter gap

The share of a company's reputation that is formed in direct experience which nobody writes down, and which is therefore invisible to a model trained on or searching the electronic record.

Time gap

The distance between what the company is now and what has been written about it. Retrieval narrows the time gap relative to the parametric layer. Retrieval does not close the encounter gap.

Parametric layer

What a language model answers from its trained weights, without searching. The model cannot reliably name the source, and the layer changes when the model is retrained.

Retrieval layer

What a language model answers after fetching pages. Sources can be audited at link level. The layer is current but selective, and it inherits the written record's tilt toward what was written down, including complaints and news.

What reputation is, and what it is not

DEFINITION. Reputation is the shared belief that a group of people holds about a company. It is built from individual encounters: directly in service situations and workplaces, indirectly through the company's own messages, and second-hand through journalists, influencers, and private conversations. It is rebuilt as new encounters land on old beliefs.

DEFINITION. Brand is not reputation. Brand is customer-centric, differentiates the offer, and is chosen by the company. Reputation covers the operating environment. Its dimensions apply whether or not the company would have chosen them. Without reputation, a brand is described here as a set of empty promises. Without a brand, reputation is described as a ticket to a match with no indication of who is playing.

INTERPRETATION. Boards lose time when the chair, the chief executive, the chief marketing officer, and the head of communications use the same word for different things. The practical instruction is to agree the meaning before the decision. The research literature does not settle the distinction, because the literature does not agree either.

Who decides

DEFINITION. The four resources a company cannot do without, following Charles Fombrun's formulation as used here, are customers, current and future employees, providers of capital, and the societies in which the company operates. No customers, no employees, no capital, or no licence means no business.

INTERPRETATION. Igor Ansoff treated stakeholders as a constraint on shareholder value. R. Edward Freeman, in the 1980s, reversed the lens: every stakeholder a company can affect can affect the company, and shared interests exceed conflicts, so the strategy is cooperation. The claim that matters for this document is Fombrun's: the company with the strongest support from the four groups is the most competitive company in its market.

The four groups are autonomous. Customers choose whom they buy from. Talented people choose where they work. Investors choose where they place capital. Societies legislate through democratic processes. The reading that governs those choices is called reputation: the stakeholders' estimate of the company's ability, manner, and willingness to produce things they value.

The value chain asserted here runs in one direction. Encounters form perceptions. Perceptions aggregate into reputation. Reputation is associated with behaviour: buying, recommending, applying, investing, trusting, granting or withholding permission. Behaviour is the business result. Reputation is not a substitute for selling a needed product at a payable price in a findable place. It is described as an additional force on the free choices of the people the company depends on.

INTERPRETATION. The asset is not in the company's vault. It is in other people's heads, where management has no direct control. That is the stated reason it is otherwise left with whoever interprets other people's thoughts for a living.

Eight European markets, spring 2026: reputation and stakeholder support among the general public

MEASURED. In every one of eight European markets measured in spring 2026, the slope of stakeholder support on reputation was above 1. The range is 1.08 in Estonia to 1.20 in France. The same model, the same quality criteria, and the same fieldwork window were used in every country. The figures concern the general public, because that is the population whose results can be published without disclosing client work. The general public is not claimed to be the most important stakeholder group. It is the context in which the other groups operate, and the population used here for an overall reputation comparison.

Total validated evaluations in the table: 77,746. Finland's basket is 60 companies; each other market's basket is 30.

What the slope means, restated. Reputation and stakeholder support are both on a 1-to-5 scale. Reputation is the statistical model built from the eight Reputation&Trust dimensions. Stakeholder support is willingness to trust, buy, recommend, apply, and invest. Each company is one point. The slope is the gradient of the line through those points. A slope of 1.18 associates a reputation move from 3.0 to 4.0 with an expected support move from about 3.0 to about 4.18.

INTERPRETATION, three points the author asks a board to take.

• The shape is not only Finnish or Nordic. It appears in France, Germany, the United Kingdom, and Estonia, in markets that differ in company size, economic structure, media, and political culture.
• The slope is above 1 everywhere in this table. A difference in perception is associated with a proportionally larger difference in stated willingness to act. The gearing is not claimed to care which way the lever is pulled.
• The table is not a league table. Differences of a few hundredths are within what samples of 30 companies can resolve. No claim is made that France is more reputation-sensitive than Denmark. The robust finding stated here is the range 1.08 to 1.20, with every market above 1.

MEASURED. Finland's own series, on a comparable basis since 2013, moved from a slope of 1.15 in 2020 to 1.18 in 2026. The step is small. It is described as consistent with a 2022 observation that the association between reputation and stakeholder behaviour had tightened rather than loosened over the previous decade.

How the 2026 eight-market figures were produced

• Design. Each market has a two-phase national sample. In phase one, 1,000 citizens, balanced by age, gender, and region, name two companies with a notably good reputation and two with a notably poor one. The 30 most-named companies form the basket. Finland's basket is 60: 30 from nominations and 30 selected to balance industries, so the Finnish series stays comparable with earlier years.
• Measurement. Every company in the basket is evaluated with the full Reputation&Trust model, eight reputation dimensions and the stakeholder-support measures, by respondents balanced by age, gender, and region.
• Validated evaluations. A response counts only if the respondent knows the company, is willing to evaluate it, and evaluates it rather than answering "don't know" throughout. The 77,746 figure is validated evaluations, not raw respondents.
• Precision. Each company is measured until it has at least 200 validated organisation evaluations. At that level the statistical margin of error of the company's overall reputation score is 0.04 to 0.06 on the 1-to-5 scale, depending on the standard deviation of responses. Reaching 200 validated evaluations takes roughly 300 to more than 3,000 respondents, depending on how well the company is known. The input varies. The precision target does not.
• Fieldwork. 13 May 2026 to 23 June 2026, simultaneously in all eight markets. Timing is therefore not an explanation for differences between markets.
• Slope. Ordinary linear regression of stakeholder support on reputation across the companies in a market: 30 observations per market, 60 for Finland.
• Not yet published at the date of this article: confidence intervals for the slopes. That is why hundredths-level differences are not read as a rank order.
• Selection note. A nomination-based basket over-represents companies with notably strong or notably weak reputations and under-represents the middle. That widens the spread of points, which makes the slope estimable. The basket is not a random sample of the corporate population.

Three channels from reputation to the income statement

INTERPRETATION. Reputation is said to reach the income statement through three channels a board already governs: enterprise risk, demand, and the cost of capital. Each channel below mixes the author's argument with the cited measurements.

Risk and regulation

Two consequences are singled out. First, talent: people seek companies with strong reputations, and the ability to attract them erodes as reputation erodes. Second, regulation: reputation crises have repeatedly preceded industry-rule changes, with payday lending, electricity distribution, private care, and gambling named as examples. Where self-regulation does not satisfy society, democratic processes are described as attending to the matter efficiently.

INTERPRETATION. Some boards have placed reputation among key non-financial indicators and given it weight in incentives, so that proposals are written with stakeholder expectations in view. The author calls this a democratic element in remuneration, and suggests that some strategic decisions later judged failures might not have been taken had reputation been in the management system at the time.

Demand and market share

MEASURED. A meta-analysis of the Nordic financial sector, 139 reputation scores from four countries in 2019 to 2023, gave a slope of 1.24, above every national general-public figure in the 2026 table. The author reads that as expected where products are similar and switching is easy.

INTERPRETATION. Where the offer is reasonably homogeneous and available, market share is said to follow reputation relative to competitors, provided price and availability are in order. A company may instead hold share and take the gain as margin, with the reputational risk that pricing power carries. Where the offer is differentiated, reputation is still said to drive demand through willingness to buy.

CITED EXTERNAL FINDING. Forrester buyer research is cited for the claim that about 90 per cent of business-to-business buyers begin with a search, and three in four complete a background check of financials, conduct, and increasingly ESG performance, before contact with the supplier. Trust is described as the attribute buyers across industries say they want most.

MEASURED. A meta-analysis of business-to-business buyer studies conducted between 2013 and 2022, using the Reputation&Trust model, gives a slope of 1.22 for buyers' trust and willingness to recommend a supplier. Impact analysis of the same data is said to put the weight on products and services, governance and transparency, dialogue, and the credibility of the leadership team.

INTERPRETATION. The line "nobody ever got fired for buying IBM" is used as a description of the mechanism: a supplier's reputation is part of the buyer's own risk management. A weak supplier reputation makes the buyer look careless. A strong one absorbs the blow if something goes wrong. Business-to-business purchasing is therefore not treated as immune.

Cost of capital

MEASURED. Reputation and Trust Analytics measures annually the reputation of every Nasdaq Helsinki Large and Mid Cap company among Finnish private investors. The meta-analysis of those studies, to 2024, shows willingness to invest in and recommend a company following reputation with a slope of 1.46.

INTERPRETATION. Institutions hold most shares, but private investors can dominate daily trading and thereby set the price at which institutional portfolios are marked. Institutions are also said to ask increasingly about the reputation of holdings. Share price sets the terms of equity finance, the strength of the hand in a share-based acquisition, and, through perceived financial health, creditworthiness and the price of debt. Reputation is not said to set the price of money alone. It is said to move it.

Slopes side by side

These slopes share the Reputation&Trust model and a 1-to-5 scale. They are not directly comparable, because the populations and the support questions differ. Every reported slope is above 1.

A crisis is not binary

DEFINITION. Reputation is not present or absent. It sits at a level at every moment. A board that does not know the level cannot know what a loss would remove.

INTERPRETATION. Inside the band that modelling would call a good reputation there is large variation in capacity to generate stakeholder support. Many companies have not established where they sit. After a crisis, the pre-crisis level is often remembered as excellent and was never verified.

The saying attributed to Warren Buffett, that reputation takes twenty years to build and five minutes to destroy, is said here to hold only where the foundations were a lie. Most corporate crises are described as beginning with accidents, operational errors, a legal but unacceptable favouring of one stakeholder, or a combination. The worst industrial catastrophes named, BP Deepwater Horizon in 2010 and Bhopal in 1984, are described as following one pattern: management economised on safety to protect margins, to look good to owners, and to secure bonuses the board had set, on the assumption that nothing had gone wrong before.

Three things are said to decide the outcome. The board can set two of them in advance. First, the physical and ethical condition of the foundations, which journalists and regulators will inspect and which cannot be repaired in the moment. Second, the level and structure of reputation on the day the storm breaks. Third, crisis handling and its communication, the only one of the three that can be practised after the event starts.

The first two are said to work through documented biases. People resist information that contradicts what they already believe (cognitive dissonance) and accept information that confirms it (confirmation bias). A company with a strong reputation and clean foundations is said to benefit from both. A company with a weak reputation and real skeletons is said to receive the opposite treatment.

CITED EXTERNAL FINDING. Rory Knight and Deborah Pretty of Oxford studied share-price reactions to corporate catastrophes and found that half the affected companies recovered to or above their pre-crisis valuation within about fifty days, while the other half did not recover within the study horizon. The companies that came through are described as having been tested in public and found sound.

Operating environment and responsibility

INTERPRETATION. Reputation is formed in a time, place, and climate of values the company does not control. The variables that matter for a reputation environment analysis are stakeholders' values, attitudes, culture, public mood, and political temperature, because these set the frame in which actions are interpreted. The frame can move faster than the business cycle. An environmental programme of two years earlier can be read as greenwashing. The analysis has to be maintained, not produced once.

EXAMPLE. When the European Union abolished roaming charges in 2017, operators' business analyses are said to have seen a threat. The reputation of the sector is said to have risen afterwards, most for those who put the lower prices at the centre of their story.

On responsibility, two external anchors are cited. Milton Friedman's 1970 position, that a company's only social responsibility is to increase its profits, was adopted in the 1980s and was later called by Jack Welch "the dumbest idea in the world": shareholder value as an outcome, not a strategy. Eccles, Ioannou, and Serafeim tracked ninety companies that adopted environmental and social policies in the early 1990s against ninety matched peers; one dollar invested in the first group in 1992 is described as worth considerably more by 2010. Whelan and colleagues at NYU Stern combined more than a thousand studies published between 2015 and 2020: good ESG management is associated in that meta-analysis with higher returns on capital, higher valuations, and lower risk, and only a handful of studies find the reverse.

INTERPRETATION. The finance literature is said not to have explained why. The explanation offered here is stakeholder theory: people cooperate with organisations they believe act sustainably and decently. The Reputation&Trust model, developed inductively before ESG was mainstream, is said to have surfaced the ESG components as reputation dimensions without anyone looking for them: environment and society under responsibility, treatment of people under workplace, openness and ethical conduct under governance, with dialogue alongside them.

Two board consequences are drawn. Responsibility is to be managed on two axes: measured performance, which is industry-specific and rests on materiality, and measured perception, which is what is said to drive behaviour. Performance without perception leaves value unrealised. Perception without performance is called greenwashing, and greenwashing is said to have short legs. Second, expectations rise continuously and companies are judged against peers, so a company that improves more slowly than expectations can see perception scores fall while it improves. The board question stated here is not whether the company does more than it did, but whether it moves faster than the expectations around it.

Reputation as a reserve against information influence

DEFINITION. An information attack targets trust. The reserve of trust named here is reputation. The reserve has to be built before it is needed.

INTERPRETATION. Cyber attacks and information operations are said to share a method: both exploit trust relationships, and both are more efficient where the attacker knows where trust is thin. A statistical impact analysis of an organisation's reputation is compared to a penetration-test report. It shows which perceptions most build or undermine the licence to operate, and which perceptions are weakly held or sharply divided. The author says boards should treat that material as sensitive.

Two misconceptions are rejected. An organisation few people know is an easier target, not a harder one, because beliefs can be created quickly where almost none exist, and once created they are hard to displace. Being well known is not itself protection. Awareness combined with a weak reputation is called the most fertile ground for an attacker, because confirmation bias makes a new negative claim believable. Awareness combined with a strong reputation is called the opposite: negative information about a trusted institution is not readily believed even when it is true. That use of cognitive dissonance is described as the only version of "reputation carries you through a crisis" that the author accepts as true.

MEASURED. Public-sector studies using the same model show a slope of 1.11 between reputation and citizen support, defined as trust, willingness to fund, and willingness to listen. That support is called the institution's licence to operate.

INTERPRETATION. Breaking is faster than building, and the attacker can wait. Resilience in the information domain is therefore placed alongside cyber security on the standing risk agenda, not only in the crisis plan.

Language models as a new third party

DEFINITION. This document does not treat a language model as a stakeholder. A model has no stake in the company, bears no risk from its success or failure, and pursues no agenda of its own, so there is nothing to negotiate with. It does have a view of many organisations, it infers the rest, and stakeholders ask it about their decisions.

Reputation was described above as the product of three kinds of encounter: direct experience, the company's own messages, and third parties who form a view and pass it on. Media and social media are third parties with a perspective the reader can weigh. A generative model is a third party of a different kind: it reads what has been written and returns a synthesis, in one voice, with authority. The consensus is then also held in the weights of a handful of models. The loop asserted here is: stakeholders consult the models, the models shape what stakeholders believe, and later models learn from what stakeholders then write.

DEFINITION. The parametric layer is what the model knows and infers from training material. It answers general questions without searching. Those are the questions in which consideration sets form, where a company earns or loses the right to be compared at all. The layer is opaque. Asked for a source, the model may invent a plausible one. It changes when the model is retrained.

DEFINITION. The retrieval layer is what the model fetches when a question names a company, asks for a comparison, or needs a current fact. It is transparent to the level of the link. Sources can be audited and, where wrong or stale, corrected. It is current but selective: which pages rank, in which language, with what authority, and how the model summarises them. It inherits the written record's tilt toward what was written down.

The boundary between the layers moves with model releases, and models decide when to search. The structure claimed to hold is that the early, general conversation often stays unsearched, and that is where reputation does its oldest work of deciding who is considered at all.

What neither layer contains is a service encounter that nobody wrote down. A complaint, a review, or a news story is in the record. An undocumented good experience is not. Retrieval narrows the time gap. It does nothing for the encounter gap. The parametric layer carries both gaps.

Exploratory 2026 model measurements

METHOD. In 2026 the full Reputation&Trust questionnaire was put to four leading models: OpenAI's model, Microsoft Copilot, Anthropic's Claude, and Google's Gemini. Synthetic agent respondents were built to mimic a given stakeholder profile. Each run was done once with search disabled and once with search enabled. Results were set against reputation measured among real people in the same group. These are exploratory measurements, not a series. A figure is valid until the next model release.

• Reputation and Trust Analytics, as its customers see it: 4.11 among customers, 3.90 in the retrieval layer, 3.67 in the parametric layer. Search closes about half the distance. Training-based knowledge closes none of it. Read as both gaps: a specialist firm well known to the people it serves and thinly written about.
• Stora Enso, as the general public sees it: about the same overall level in the models as among people, but a very different structure across the eight dimensions. Read as the time gap: the models reach a similar verdict about a company in transformation by weighing the dimensions of the company it used to be.
• Masku, as the general public sees it: a markedly weaker reputation in the models than among people. Read as the encounter gap: a retailer's standing is made in the shop and rarely written down except in complaint.

INTERPRETATION. The more a company's competitiveness is built in direct encounters, and the faster it has changed, the more certainly the models underrate it. The mirror image is also stated: a company whose written record is better than its daily conduct is lent reputation by the models until the record catches up.

For information-influence risk, the layers differ. A narrative in the retrieval layer is traceable to a source and can be contested. A narrative absorbed into the parametric layer is repeated as the model's own knowledge, its origin cannot be proved from the answer, and it cannot be corrected until the training material has changed enough.

The governance split stated here: the retrieval layer is managed operationally, as an extension of media monitoring. The parametric layer is measured the way a stakeholder is measured, by asking repeatedly on a comparable scale, and influenced slowly through what the company puts on the record. The encounter layer is beyond both until the company documents what it now leaves undocumented. The task is described as making real performance exist electronically, so that the consensus forms from what the company is rather than from what happens to have been written about it.

Reputation-management maturity model

This model is distilled from practice, not from a research programme. It describes the structure of management, not the content of reputation. A mature structure does not guarantee success. It is said to guarantee a systematic grip, traceability, and continuity when people change. No company is claimed to sit exactly at level 4.

Diagnostic use: if, when the board says reputation, management hears fear, the company is at level 1. The way out described here is the same in the cases the author has seen: model the company's actual reputation statistically, which produces both a working definition and a number. Once it can be discussed as business impact, the conversation spreads, and maturity follows the conversation.

OBSERVATION. Companies with long-standing excellent reputations have often spent most of their history at level 1, steered by owners who cared, and have typically been early adopters of systematic tools once tools existed. Where owners care only about the next quarter, long-term reputation thinking is said to carry no weight. The private care sector's loss of legitimacy in Finland is named as a case of both kinds of company sharing one industry's fate.

One situational picture

INTERPRETATION. An executive team cannot develop a reputation it perceives differently. Customer satisfaction, employer brand, personnel surveys, political mapping, sustainability studies, and brand tracking are each useful and none is commensurable with the others. Reputation is said to require its own instrument, because no single-purpose metric captures the holistic view of a company as an economic entity, an employer, a member of society, a producer, and a moral actor at once.

ILLUSTRATION, composite, not a named client. A large business-to-business company, called Jätti Oyj in the source article, is behind its sector on the general-public benchmark. The general public is the starting point because it is the context and the cheapest comparable benchmark, not because it is the most important stakeholder for a business-to-business firm. Objections about who was asked and whether one article moved the score are answered by the method: same standard, same time, margin of error 0.04, and the previous year showed the same gap. The structure of the score is highly profitable and financially stable, and a poor listener, with weak scores on governance, responsibility, and value for money. The next question is operational versus communications: if perceptions match performance, the problem is not communication; if performance exceeds belief, communication can close the gap. The environment is the same for peers. The benchmark is peers.

The eight dimensions are not equally important. Weight depends on industry, current standing, business context, and what drives the stakeholders. Impact analysis, correlation and regression against stakeholder-support measures, is what directs scarce attention. In the Jätti illustration, external employer image had an unusually strong effect on overall support, which turned a communications question into a people-management programme.

Cadence recommended here: one key stakeholder group per quarter, discussed against the previous results, rather than every group at once. Within a year the company then has customers, employees, investors, and political decision-makers, and the author reports that drivers usually rhyme across groups.

Targets and remuneration

There is no single numerical target that fits every company. Targets are subordinate to strategy. A target that is never fixed cannot be met or missed.

Defenders already hold a strong reputation, often the strongest in their industry. From inside, that reputation rests on historical performance that has to be lived out every day. Defender targets should be anchored to the movement of the market or the industry, not to the company's own score. If employer image is improving across the market, a flat score is lost ground. If the market falls sharply and the company dips slightly, that is gained ground.

Attackers search the impact analysis for dimensions where they can move. MEASURED DISTRIBUTION. Among companies that have actively set out to improve their measured reputation, an increase of 0.15 on the 1-to-5 scale over a year places the company in the top quarter of those that succeeded. The same kind of distribution exists for individual dimensions, so a dialogue spearhead can have a dimension-specific target.

INTERPRETATION. Making reputation a remuneration criterion is the most direct board lever in this document. It is described as long-term governance, not a communications initiative. Managers change. The company, its reputation, and its stakeholders remain. Continuous measurement and linked remuneration are the board's means of continuity on that side. The organisation still has to produce the ideas. The board's contribution at that point is pressure, resources, and a signal that success on the measure will be rewarded.

Board checklist

Eight questions. A company that can answer all eight is, on this document's terms, managing the intangible asset. A company that cannot answer the first is not.

1. Do we know the level and structure of our reputation, measured on a comparable basis, among the general public and among each key stakeholder group?
2. Do we know where we sit relative to our closest competitors, and has the gap moved over the past three measurements?
3. Do we know which reputation dimensions actually drive stakeholder support for this company, rather than assuming they are equally important?
4. Do we know what the leading language models say about the company, on the same scale, in both the parametric layer and the retrieval layer, and where that diverges from what people say?
5. Who holds responsibility for reputation, and does that person have the mandate and the resources that responsibility for competitiveness and enterprise risk requires?
6. Have we set a numerical reputation target anchored to the market's movement or to the empirical distribution of achievable change, and is it part of management remuneration?
7. Does our environment analysis track changes in stakeholder values and expectations, especially on responsibility, and is it updated often enough to matter?
8. Are our physical and ethical foundations in a condition we would be comfortable having inspected in public tomorrow?

Prior evidence cited, not re-measured here

• Aon's Global Risk Management Survey has kept damage to reputation among the top ten enterprise risks for more than a decade. In one edition, risk professionals from more than sixty countries rated it the single greatest risk out of more than fifty.
• Richard Hall's 1992 study found chief executives ranking reputation the most important intangible resource for sustainable competitive advantage.
• Charles Fombrun and Jonathan Low showed strong-reputation companies outperforming the S&P 500 over five years.
• Reputation Dividend has put reputation at 38 per cent of the value of the FTSE 100 and 250, roughly a trillion pounds. The method may be disputed. The order of magnitude is what the author treats as undisputed.

Questions and answers

What is reputation in this document?

Reputation is the shared belief a group of people holds about a company. It is built from encounters, direct and indirect, and it is not the same thing as brand. Brand differentiates the offer and can be chosen. Reputation covers the operating environment and is not chosen dimension by dimension.

What is the Reputation&Trust model?

It is an eight-dimension model used since 2013 by Reputation and Trust Analytics. The dimensions are financial performance, leadership, innovation, products and services, dialogue with stakeholders, workplace, responsibility, and governance. Numbers in this document that come from the author's own measurements rest on that model and a 1-to-5 scale.

What did the 2026 eight-market study find?

Fieldwork from 13 May 2026 to 23 June 2026, general public, eight European markets, 77,746 validated organisation evaluations. The slope of stakeholder support on reputation was above 1 in every market: France 1.20, Finland 1.18, Sweden 1.17, Germany 1.12, United Kingdom 1.11, Norway 1.10, Denmark 1.10, Estonia 1.08. The table is not a ranking of markets. Confidence intervals for the slopes were not published with the article.

What does a slope of 1.18 mean?

On a 1-to-5 scale, a company whose reputation moves from 3.0 to 4.0 has an expected stakeholder-support move from about 3.0 to about 4.18. Stakeholder support means stated willingness to trust, buy, recommend, apply, and invest.

Are the slopes causal?

They are regression coefficients across companies. This document describes a directional value chain from encounters to perceptions to reputation to behaviour, and it reports the association. It does not report a controlled experiment that isolates causation.

Why is the company basket not a random sample?

Phase one asks 1,000 citizens to name companies with notably good or notably poor reputations. The basket over-represents the extremes and under-represents the middle. That widens the spread and makes the slope estimable. It is a limitation, stated in the methods.

What is the encounter gap?

The encounter gap is the part of reputation formed in direct experience that leaves no electronic trace. A language model cannot see it. A sincere greeting at a counter is the article's example: it marks the customer and does not mark the record.

What is the time gap?

The time gap is how far the company has moved from what has been written about it. Retrieval narrows it, because retrieval reads the current record. The parametric layer carries it until retraining.

What did the 2026 language-model measurements show for the author's own company?

Reputation and Trust Analytics scored 4.11 among its customers, 3.90 when leading models searched before answering, and 3.67 when they answered from what they already knew. The models were OpenAI's model, Microsoft Copilot, Anthropic's Claude, and Google's Gemini. The measurement is exploratory and expires at the next model release.

Is a language model a stakeholder?

Not in the sense used here. It has no stake, bears no risk, and has no agenda to negotiate with. It is a new kind of third party that synthesises the written record and that stakeholders consult.

What should a board measure?

The level and structure of reputation among the general public and key stakeholder groups; position versus competitors over at least three measurements; which dimensions drive support; what leading models say on the same scale in both layers; who holds the mandate; a numerical target anchored to market movement or to the distribution of achieved change, linked to remuneration; an environment analysis of stakeholder expectations; and whether the physical and ethical foundations would survive a public inspection tomorrow.

What improvement counts as top-quarter?

Among companies that actively set out to improve measured reputation, an increase of 0.15 on the 1-to-5 scale over a year is the top quarter of those that succeeded, as stated in this document.

Limits of this document

• The 2026 market slopes are cross-company associations, not experimental effects.
• Baskets are nomination-based, not random samples of all companies.
• Slope confidence intervals were not published with the article. Hundredths-level differences between markets are not a ranking.
• Slopes from different populations and different support questions are not directly comparable. They are reported together only because each is above 1.
• The language-model scores are exploratory, from synthetic agent respondents, and are valid only until the next model release.
• External citations (Aon, Hall 1992, Fombrun and Low, Reputation Dividend, Forrester, Knight and Pretty, Eccles Ioannou and Serafeim, NYU Stern ESG meta-analysis) are cited, not re-analysed here.
• The maturity model is from practice, not from a validation study.
• Jätti Oyj is a composite illustration.

How to cite

Riku Ruokolahti. "Reputation in the age of AI: definitions, measurements, and board governance." Reputation and Trust Analytics, 9 October 2026. Eight-market general-public slopes fielded 13 May 2026 to 23 June 2026; AI-layer scores described as exploratory.