What Positioning Means in 2026

Positioning has always been an argument about context.

Ries and Trout taught marketers that products compete for a place in the prospect’s mind, and that the mind is crowded, selective and economical. The category in which a product is placed determines the expectations brought to it. The first credible claim can become a reference point against which later claims are judged. A company does not enter an empty market with a neutral object. It enters an existing structure of memory, language and alternatives.

April Dunford made this operational for B2B technology. Start with the alternatives a customer would use if the product did not exist. Identify the capabilities that are genuinely different. Translate those differences into value. Determine who cares intensely. Select a market frame that makes those strengths obvious. Her work remains one of the most useful corrections to the vague, fill-in-the-blanks positioning statements that dominated marketing practice.

None of this has stopped being true.

But the environment in which positioning is interpreted has changed. The buyer no longer encounters a company’s argument mainly through a sequence of pages, ads, analyst reports and seller conversations selected by the company. Research is increasingly assembled by AI systems that retrieve fragments, compare claims, infer categories, summarise reviews, generate shortlists and answer questions in language the buyer invents in the moment.

Google describes AI Mode as helping people “think through considerations and narrow down products” by combining model capabilities with its product graph. OpenAI’s shopping research asks clarifying questions, researches sources, compares constraints and returns a personalised buyer’s guide. These examples are consumer-facing, but the interaction pattern is already familiar in B2B. A buyer asks a system to define the problem, identify approaches, compare vendors, expose trade-offs, draft requirements and prepare questions for a sales call.

The buyer journey has acquired an interpreter. This interpreter is tireless, fast and occasionally has the social confidence of a man explaining your own job to you at a wedding.

Positioning in 2026 must therefore do two things at once. It must still establish a useful human mental frame. It must also create a coherent evidence structure from which machines can reconstruct that frame without flattening the product into generic category attributes.

This does not mean writing for robots. It means recognising that the position is now executed across a distributed information environment in which human and machine interpretation interact.

1. What the classics got right

It is fashionable to declare old frameworks obsolete whenever the interface changes. That is usually an attempt to manufacture novelty.

Ries and Trout remain useful because cognitive constraint remains real. A buyer cannot evaluate every product from first principles. Categories reduce the cost of understanding. Positions create retrieval shortcuts. Relative claims are easier to process than isolated claims. “The first adaptive fraud-detection system for issuers” gives the mind more structure than a paragraph of capabilities.

Dunford remains useful because she starts with market reality rather than company aspiration. Her current description of Obviously Awesome still centres competitive alternatives, differentiated value, best-fit customers and the market context in which those strengths matter. That sequence forces teams to confront the difference between what they built and why a customer would choose it.

Three principles survive intact.

Positioning is relative

“Easy to use” exists only in comparison with an alternative, for a specific person attempting a specific job under specific conditions. “Enterprise ready” demands more than a permissions page and a photograph of three serious people near a glass wall. The useful question is whether the product is more governable than the current approach for an organisation with a particular risk model.

AI systems do not remove relativity. They intensify it. A comparison answer is explicitly relational. The system looks for difference, fit, constraints and evidence. Generic superlatives become even less useful when a model can place six vendors in a table and discover that all claim to be intelligent, unified and seamless.

Positioning is selective

Every strong position excludes. It chooses the value that matters, the customer who cares and the category assumptions worth invoking. The product may have many capabilities. The position determines which capabilities become commercially legible.

This selectivity is more important when content generation is cheap. AI enables companies to produce pages for every segment and use case. Without a stable position, that abundance creates semantic sprawl. Each page is locally plausible, while the market receives no durable answer to what the company is and why it wins. It is possible to own a great many matching beige towels and still have no idea where the bathroom is.

Positioning is strategic

Positioning influences product priorities, qualification, packaging, pricing, partnerships and proof. A company positioned as infrastructure makes different roadmap and trust commitments from one positioned as a productivity tool. A platform position creates expectations about breadth, integration, governance and durability. Those expectations become liabilities if the company adopts the label only because it sounds larger.

The classic foundations remain sound. Trouble begins when a company treats the positioning decision as complete once a workshop has produced a message document.

2. The buyer journey changed interface

Traditional journey maps present awareness, consideration, evaluation and purchase as a sequence. Sophisticated teams have always known the sequence is messy, but the company could still design content and touchpoints around a broadly progressive flow.

AI makes the journey more recursive and more compressed.

A buyer can begin with an implementation question, move backwards into category definition, request a shortlist, discover an unknown risk, revise the requirements and ask for a business case in one session. The system can maintain context across these moves. It can convert vague intent into criteria before a vendor knows the account exists.

The buyer can ask:

• What are the main ways to solve this problem?
• Which approach works for a regulated multi-entity organisation?
• Compare the trade-offs between building internally and buying.
• Which vendors support data residency and explainable decisions?
• What do customers complain about after implementation?
• Draft an RFP that would separate modern platforms from legacy suites.
• Which claims should I challenge in a demo?

Each answer can shape the category, criteria and shortlist. The company’s homepage may appear only as one source among product documentation, customer reviews, partner pages, public pricing, community discussions, job descriptions and competitor comparisons.

The implication is not that websites no longer matter. Websites become canonical evidence surfaces rather than the only stage on which the story is told.

Google’s product announcements make the direction explicit: conversational systems combine product data, web information and inferred preferences to support inspiration, consideration and narrowing. OpenAI describes systems that review sources, apply constraints and produce a personalised guide. In 2026, Google is also publicly discussing agentic commerce infrastructure and merchant visibility inside AI surfaces. B2B will move more slowly because the objects are complex and the decisions are organisational, but the interface logic travels.

The old journey asked, “Which page should the buyer see next?”

The new journey asks, “Can the market reconstruct our position accurately from whatever evidence surface the buyer or agent consults next?”

3. Machine mediation changes the failure modes

Human buyers have always misinterpreted positioning. Machine mediation creates new ways for that misinterpretation to scale.

Category collapse

Models are trained and retrieved over existing language. If a company describes itself with broad, common phrases, the system is likely to map it into the nearest established category. A genuinely different architecture can disappear into a familiar label because the public evidence does not make the distinction resolvable.

This is category collapse: the market sees a novel product as another member of the class it is trying to replace.

The response is not to invent an opaque category name and repeat it. A new frame must be connected to known alternatives, concrete capability and observable value. Machines need bridges just as humans do.

Attribute flattening

Comparison systems convert narratives into fields. Integrations, deployment, pricing, security, use cases, customer size. If the position depends on a system-level advantage, but the evidence exists only as a high-level slogan, the comparison collapses to feature presence.

A fraud platform whose advantage lies in adaptive behavioural models may be represented as “AI: yes” alongside a rules engine with a chatbot. A construction platform whose advantage lies in cross-company workflow continuity may be compared by module count. The differentiated mechanism vanishes.

Evidence substitution

When first-party evidence is unclear, a system may rely on third-party summaries, old pages, reviews or competitor descriptions. The market fills the vacuum. This has always happened, but AI can synthesise the substituted account into a confident answer.

Positioning must therefore include evidence governance: which sources are canonical, how claims are supported, how changes propagate and how obsolete descriptions are retired.

Persona averaging

Generative systems are good at producing plausible messages for a persona. That can tempt teams to create countless tailored variants. But persona personalisation without a stable strategic position often averages the product into generic relevance. Every audience hears that the product improves their outcomes. Nobody learns what makes the approach distinct.

Temporal inconsistency

Products change faster than market memory. Old documentation, launch posts, pricing pages and partner descriptions remain retrievable. A model may encounter several realities at once. If the company cannot express what changed and when, the generated answer may blend them.

Positioning in 2026 must be temporal. It needs version, effective date and change logic, particularly for AI products whose capabilities and boundaries evolve quickly.

The modern positioning risk is confident reconstruction from inconsistent evidence.

4. A definition for 2026

Here is the definition I use:

Positioning is the governed system of market context, differentiated mechanism, value, proof and boundaries that enables humans and machines to understand when a product is the right choice, against which alternatives, and why.

Every word carries work.

Governed system means the position has ownership, structure, versions and operational consequences. It is not a document abandoned after approval.

Market context is the frame that activates useful assumptions. It explains what is changing and which existing approach has become insufficient.

Differentiated mechanism is how the product creates a different result. This is the missing layer in much AI-era messaging. Outcomes alone are easy to imitate. Features alone are easy to flatten. The mechanism connects capability to consequence.

Value is the operational, financial, strategic or emotional consequence for a specific customer condition.

Proof makes the claim resolvable. It can include customer outcomes, product behaviour, benchmark methodology, architecture, implementation evidence and credible third-party validation.

Boundaries state where the product is not the right choice, where human oversight remains necessary and which conditions are required for value. Boundaries increase trust and improve machine interpretation because they create discriminating criteria.

Humans and machines acknowledges the dual audience without confusing their needs. Humans need meaning, confidence and organisational permission. Machines need accessible, consistent and attributable evidence.

When is as important as who. Customer fit is often conditional: a trigger, maturity state, architecture or forcing event makes the position relevant.

5. The positioning stack

A single positioning statement cannot carry this work. I use a stack with seven connected layers.

Layer 1: market change

What changed outside the product? A regulatory shift, technical possibility, buyer expectation, cost structure, risk or workflow transition. The change creates a reason to reconsider the current approach.

Good market-change language is specific enough to be challenged. “The world is changing faster than ever” is not a strategic narrative. “Real-time payments removed the review window on which batch fraud controls depended” is.

Layer 2: competitive reality

What does the customer do now? Include internal processes, adjacent products, outsourcing and delay. The real alternative is often the existing operating arrangement plus the political cost of changing it. Every spreadsheet that “will do for now” has a small constituency and, somewhere, a person called Martin who knows all the macros.

In an AI-mediated journey, alternatives should be explicit on first-party surfaces. Companies that refuse to discuss alternatives leave the comparison to sources with weaker knowledge or different incentives.

Layer 3: differentiated mechanism

What does the product do differently at the level of cause?

Examples:

• Uses adaptive behavioural profiles rather than static threshold rules.
• Preserves project context across company boundaries rather than moving documents between isolated repositories.
• Converts market events into governed field actions rather than issuing undifferentiated alerts.

Mechanism is the smallest technically honest explanation of why a different outcome occurs. Architecture detail earns its place when it explains cause. Otherwise it is the verbal equivalent of opening the boiler cupboard during a house viewing.

Layer 4: value consequence

Translate the mechanism through the customer’s operating model. A model reduces false positives; investigators recover capacity; genuine risk receives more attention; customer interruption falls; control improves without proportional headcount.

Value should be represented as a causal chain, not a bag of benefits.

Layer 5: customer condition

Who experiences the consequence intensely, can realise the mechanism and has permission to change? Firmographics may help locate the account, but the condition explains fit.

Layer 6: proof and trust

What evidence allows each stakeholder to accept the chain? The economic buyer may need quantified consequence. The technical evaluator needs architecture and integration behaviour. Risk needs governance and boundaries. Users need workflow evidence.

Layer 7: expression

Only now do we reach messaging: category language, homepage copy, pitch narrative, comparison pages, documentation, analyst briefings, product UX and sales talk tracks.

Expression is a set of compiled outputs from the positioning stack. If every team edits outputs without reference to the stack, drift is inevitable.

6. Evidence architecture is part of positioning

For decades, proof was treated as a content type. The company developed case studies, testimonials and analyst quotes after the message was agreed. In an AI-mediated journey, proof must be designed as architecture.

The objective is resolvability: a buyer or system should be able to determine whether a claim is supported, under what conditions and by what evidence.

Use claim objects

For each strategic claim, maintain:

• The exact claim and approved variants.
• The mechanism that supports it.
• The customer and use-case conditions.
• The evidence source.
• The owner and review date.
• The confidence level.
• The boundaries and exceptions.
• The surfaces on which the claim appears.

This sounds like content operations because it is. Positioning without claim governance becomes stale at the speed of product change.

Separate evidence classes

Not all proof does the same job.

• Outcome evidence shows that value occurred.
• Mechanism evidence shows why the product can produce it.
• Adoption evidence shows that real organisations can implement and use it.
• Trust evidence shows governance, security, reliability and accountability.
• Comparative evidence shows difference against alternatives.
• Boundary evidence shows where the claim does not apply.

A customer logo is weak evidence if nobody can tell what changed. A benchmark is weak if the methodology does not resemble the buyer’s environment. A technical explanation is weak if it never reaches a business consequence.

Build canonical surfaces

The homepage cannot contain everything. Create durable pages for the category, approach, architecture, security, implementation, comparisons, use cases and customer evidence. Use consistent terminology and links between them. Structured data can help machines identify the entities and relationships, but markup cannot rescue incoherent substance.

Publish change

When the product or position changes, explain the transition. Release notes, migration guides and updated documentation should make temporal state clear. Do not simply overwrite the old claim if third-party references depend on it.

Make expertise attributable

AI systems evaluate source quality imperfectly, but clear authorship, relevant experience and original analysis improve the information environment for humans and machines. Anonymous corporate prose is difficult to assess. A named expert with a visible record and defensible argument creates a stronger trust path.

This is why thought leadership cannot be separated from positioning. Original thinking supplies category language, causal explanation and evidence that generic pages cannot.

7. Positioning becomes a live operation

The old cadence was episodic: reposition during a rebrand, product launch or leadership change. The 2026 cadence is continuous but controlled.

Continuous does not mean constantly rewriting the homepage. It means monitoring whether the assumptions beneath the position remain true.

Track four kinds of signal.

Market signals

New regulation, buyer vocabulary, technology shifts, category consolidation and changes in how the problem is framed.

Competitive signals

Not every feature announcement matters. Look for changes that alter the comparison: pricing model, distribution, integration, proof, market frame, target segment or product architecture.

Buyer signals

Questions asked before contact, criteria appearing in RFPs, objections, reasons for no decision, internal business-case language and the sources buyers trust.

Product signals

Capabilities, limitations, implementation patterns, adoption behaviour and realised outcomes. AI products especially require close coordination because model behaviour and governance can change the truth of a claim.

Route these signals through a decision cadence. Some update expression. Some challenge the mechanism or target condition. Some are noise. The position should be stable enough to accumulate memory and flexible enough to respond when the underlying reality changes.

The operating team is cross-functional. Product marketing may own the system, but product, sales, customer success, legal, security and leadership contribute evidence and accept consequences. If positioning is “marketing’s wording”, other functions will bypass it as soon as a deal or launch creates pressure.

Use version control principles:

• A canonical current position.
• A record of material changes and rationale.
• Named owners and approvers.
• Downstream dependencies: pages, decks, prompts, enablement, product copy.
• A deprecation process for obsolete claims.
• Tests: whether target buyers understand, believe and repeat the intended frame.

The metaphor is useful because positioning now resembles a shared dependency. A breaking change propagates.

8. A modern positioning process

The workshop remains valuable, but it is the synthesis point, not the beginning.

Step 1: reconstruct buying reality

Review wins, losses, stalls and churn. Interview customers across roles. Examine search and AI referral questions where available. Ask buyers which alternatives they considered, how criteria formed, what they asked AI tools, which sources changed their view and what they needed to explain internally.

Do not ask, “What do you value about us?” and accept the first answer. Reconstruct the decision.

Step 2: map the evidence environment

Search the category as a buyer would. Ask major AI systems to define it, compare approaches and describe the company. Record which sources they cite, where the account is wrong and which distinctions disappear. You are diagnosing public meaning. Rankings are merely one symptom, rather like checking a patient’s height because it is easy to measure.

Step 3: decide the strategic stack

Agree market change, alternatives, mechanism, value consequence, customer condition and proof. Make the exclusions explicit. Test whether the category frame activates helpful assumptions without creating promises the product cannot meet.

Step 4: build the narrative and claim graph

Translate the stack into a causal story. Define claim objects and evidence. Connect stakeholder value stories to the same strategic spine. Identify which claims require new research, product instrumentation or customer proof.

Step 5: compile by surface

Produce the homepage, pitch, category page, comparisons, product documentation, analyst narrative, sales discovery and AI-answerable FAQ. Do not paste the same copy everywhere. Preserve meaning while adapting the expression to the decision at that surface.

Step 6: test comprehension and retrieval

With target buyers, test what they think the product is, who it is for, why it differs and what they would compare it with. In AI systems, test whether the category, mechanism, conditions and proof are reconstructed accurately. Record the failure pattern alongside preference.

Step 7: install governance

Assign owners, review cadence, signal inputs and downstream update rules. Positioning work is complete only when the organisation can keep it true.

9. What becomes more human, not less

If machines retrieve, summarise and compare, it is tempting to conclude that positioning becomes a data-structuring discipline. That conclusion is incomplete.

The more competent the machine becomes at processing available information, the more valuable genuinely discriminating judgment becomes.

AI can summarise customer interviews. It cannot decide which uncomfortable trade-off the company should make without being given a strategic objective and accepting no consequence. It can generate category options. It does not carry the organisational courage required to reject the category investors prefer. It can produce twenty value propositions. It does not know which promise the product and operating model should be built to keep.

Human positioning judgment is concentrated in four acts.

Choosing the frame

Many frames can be defensible. The strategist chooses the one that connects market truth, product advantage and company ambition without creating fatal expectations.

Naming the mechanism

The differentiated mechanism often exists across product architecture, customer workflow and commercial outcome. Seeing it requires technical curiosity and market empathy at once.

Making the exclusion

Positioning becomes powerful when the company says, “This is not for everyone, and this is the trade-off we make.” AI is excellent at inclusion and variation. Strategy requires refusal.

Creating belief

Organisations do not change because a comparison table is accurate. People need a story that makes the current approach feel untenable, the new approach coherent and the act of change defensible. Narrative remains a human coordination technology.

AI makes average expression abundant. It does not make a consequential position less rare.

The work of Ries and Trout taught us that context shapes perception. Dunford gave B2B teams a practical method for connecting alternatives, difference, value and fit. The AI-enabled journey does not invalidate either. It adds a new execution environment.

In 2026, a position must live in the buyer’s mind, the seller’s conversation, the product’s behaviour and the machine’s reconstruction. It must be selective enough to mean something, structured enough to travel, evidenced enough to resolve and governed enough to remain true.

The resulting work reaches far beyond copywriting. Positioning becomes a company operating discipline, complete with owners, dependencies and the occasional awkward conversation that proves a real choice has been made.