DEV Community: EntityRise

AI visibility for B2B companies: the complete guide

EntityRise — Thu, 15 Oct 2026 14:00:18 +0000

By Kamil Kępiński, founder of EntityRise

What AI visibility means for B2B companies, why buying decisions now start inside AI answers, and the five levers that decide whether AI systems recommend you or your competitors.

AI visibility is how often, how accurately and how prominently AI assistants such as ChatGPT, Claude, Gemini, Grok and Perplexity mention, describe and recommend your company when buyers ask questions in your market. For B2B companies this is no longer a side channel: buying committees now ask AI assistants for shortlists, comparisons and validation before they ever reach your website, and companies that AI systems skip are eliminated before the first sales conversation.

B2B buying now starts inside AI answers

A typical B2B purchase starts with questions, not with vendors: "how do we solve X", "best tools for Y", "Vendor A vs Vendor B". Those questions used to go to Google, where you competed for ten blue links. That has measurably changed: 51% of B2B software buyers now start research with an AI chatbot more often than with Google, and one in three has purchased from a vendor they first encountered inside an AI answer (G2, survey of 1,076 B2B software buyers and decision-makers, March 2026). Inside an AI answer there are usually two to five named companies - and the assistant explains, compares and recommends them in one message. If your company is absent from that answer, the buyer does not scroll to find you; there is nothing to scroll. This is why AI visibility behaves like distribution, not like decoration: being in the answer is the whole game.

What AI systems get wrong about B2B companies

Four failure patterns repeat across B2B audits of ChatGPT, Claude, Gemini, Grok and Perplexity. First, absence: the model simply does not recall the company for its own category. Second, misdescription: the company is mentioned but framed with the wrong category, outdated positioning or a wrong flagship product. Third, competitor substitution: the assistant answers the buyer's exact question with a rival's name because the rival's evidence is easier to verify. Fourth, citation gaps: in live AI search, answers cite third-party sources that never mention you, so even a web-connected assistant cannot bring you into the answer. Each pattern has a different fix, which is why measurement must separate them instead of producing one blended score.

How to measure AI visibility before you change anything

Measurement means running the questions your buyers actually ask - repeatedly, across several AI providers - and recording whether you are absent, mentioned or recommended, how you are described, which competitors appear and which sources are cited. One chat screenshot is an anecdote; repeated, recorded runs are a baseline you can compare against after changes. The full process EntityRise uses, including confidence labels and disclosed limitations, is public on the methodology page. The AI Visibility Audit shows what this starting measurement includes. Fix nothing until you have this baseline: without it, you cannot prove any improvement happened.

The five levers that move AI visibility

Machine-readable clarity.Server-rendered content, clean structure, schema markup and consistent entity naming so models can parse who you are and what you sell.

Category and positioning language.One clear category sentence, repeated consistently across your site and external profiles - models recommend companies they can classify.

Buyer-intent content.Pages that answer the comparison, pricing, use-case and "best for" questions buyers ask AI assistants, phrased the way the questions are phrased.

Verifiable evidence.Named customers, concrete numbers, case studies and third-party proof - AI systems favour claims they can corroborate in independent sources.

External authority and citations.Presence in the directories, review sites, industry articles and databases that AI search engines actually cite for your category.

None of these is a trick, and no lever works alone; the pattern that wins is coordinated changes measured before and after.

Frequently asked questions

Is AI visibility just SEO with a new name?

No. SEO optimises for ranked lists of links; AI visibility covers what models say from memory and what AI search engines cite in composed answers. The disciplines overlap on technical foundations but are measured completely differently.

How long does it take to improve?

Live AI-search visibility can shift within weeks of publishing better evidence and fixing accessibility, because it reflects the current web. Model-memory visibility moves more slowly, following training cycles. A serious programme measures both separately.

Can an agency guarantee AI rankings?

No one controls independent AI providers, so treat guarantees of placement with caution. What can be guaranteed is the work, the transparency and a before/after remeasurement - which is exactly what EntityRise commits to.

Where should a B2B company start?

With a measurement. Then fix the highest-impact gaps - usually entity clarity and buyer-intent evidence - and remeasure the same questions on the same providers.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

About EntityRise. EntityRise, by Smophy Labs Inc., measures how ChatGPT, Claude, Gemini, Grok and Perplexity describe, cite and recommend your company, then implements the fixes: technical, content and authority work. Every audit asks 20 buyer questions across five AI providers, seals the question set with a SHA-256 hash and records the model versions, so every change is measured against evidence, not opinion. Start with the AI Visibility Audit for $49 and see how AI assistants present your brand to buyers today. Run your AI Visibility Audit.

AI visibility vs SEO: what transfers and what doesn't

EntityRise — Wed, 14 Oct 2026 18:00:55 +0000

By Kamil Kępiński, founder of EntityRise

AI visibility and SEO overlap on foundations and diverge everywhere else. What transfers from SEO and why ranking #1 doesn't mean AI recommends you.

AI visibility and SEO are related disciplines that produce different outcomes: SEO earns positions in ranked lists of links, while AI visibility earns mentions, accurate descriptions and recommendations inside composed AI answers. A company can rank #1 on Google for its category and still be absent when buyers ask ChatGPT, Claude, Gemini, Grok or Perplexity for a shortlist - and the reverse happens too. Understanding what transfers between the two saves budget in both.

What transfers from SEO

Three SEO investments carry over almost fully. Technical hygiene: crawlable, server-rendered, well-structured pages help AI systems exactly as they help Googlebot. Genuine authority: the coverage, reviews and third-party mentions that earn ranking signals are often the same sources AI engines cite. And question-shaped content: pages built around real buyer questions feed both featured snippets and AI answers. If your SEO was built on substance, a meaningful share of the AI-visibility foundation already exists.

What does not transfer

The divergences are where budgets get wasted. Keyword positions do not transfer: AI answers name two to five companies, so being result #6 rounds to invisible. Keyword optimisation does not transfer: models classify entities and verify claims; they do not count keyword densities. Link-building metrics do not transfer one-to-one: engines cite sources they trust for a question, which is not the same list as high-DR backlinks. And SEO reporting does not transfer: rank trackers say nothing about whether ChatGPT describes your product accurately or recommends your competitor. Measuring AI visibility needs its own instrument - repeated buyer questions across providers, recorded and compared. See what the AI Visibility Audit measures.

Do you need both SEO and AI visibility?

For most B2B companies, yes - they cover different moments of the same buyer journey. Search still carries research traffic; AI assistants increasingly own the shortlist moment. The practical question is sequencing: if AI answers already recommend competitors by name while your rankings hold, the shortlist moment is where revenue is leaking now. The GEO vs AEO guide covers how the optimisation disciplines relate; the budget logic is simpler: fund the layer where buyers currently cannot find you.

Frequently asked questions

Will good SEO automatically give me AI visibility?

No - it helps the foundations but does not guarantee presence in answers. The overlap is real and partial, which is why measurement, not assumption, should decide where you invest.

Does AI visibility work hurt SEO?

Properly done, no. Entity clarity, verifiable evidence and earned citations are the substance Google's own systems reward. The two programmes share most of their raw material.

Should my SEO agency handle AI visibility?

Ask them two questions: how do they measure AI answers before and after, and across which providers. If there is no measurement instrument, there is no way to know the work worked.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

AI citations explained: how assistants choose their sources

EntityRise — Wed, 14 Oct 2026 14:00:52 +0000

By Kamil Kępiński, founder of EntityRise

What AI citations are, how assistants choose which sources to cite, why cited is not the same as ranked, and how to become citable.

An AI citation is a source an assistant references when composing a live answer - the linked footnotes in Perplexity, the source chips in ChatGPT search, the references in Gemini. Citations decide who can enter a live AI answer at all: if no cited source mentions your company, the answer cannot recommend you, no matter how good your website is. That makes citations the supply chain of AI visibility - and like any supply chain, it can be mapped, measured and improved.

How engines pick what to cite

The selection happens in two steps. First retrieval: the engine runs searches and fetches a small set of pages it considers relevant and trustworthy for the question - often comparison articles, review platforms, documentation, industry databases and news. Then synthesis: the model composes an answer grounded in those pages and attributes claims to them. Two properties follow. The cited set is small - a handful of sources carries the whole answer. And it is question-specific - the sources cited for "best X for enterprises" differ from "X pricing", so citation presence must be checked per question, not per brand.

Cited is not the same as ranked

Engines do not simply cite the top of Google. Measured answers regularly cite pages that rank modestly while skipping pages that rank high, because retrieval optimises for answerability: pages that directly, verifiably answer the question in machine-readable form. A dense comparison table on a mid-authority blog often beats a high-authority homepage that answers nothing. This is good news for challengers: the citation game rewards content quality and verifiability more than domain size.

How to become citable

Three moves, in order. Map the citations: run your buyers' questions across providers repeatedly and record which domains and pages get cited - that list is your real competitive terrain. Earn presence in third-party sources: accurate listings and reviews on the platforms that appear in your map, and coverage in the articles engines keep returning to. Publish answerable pages yourself: direct, structured, verifiable answers to the questions - because your own site can be a cited source, and for methodology, pricing and product facts it should be the cited source.

Frequently asked questions

Do all assistants cite the same sources?

No - overlap exists but each engine has its own retrieval habits, which is why citation measurement runs across ChatGPT, Claude, Gemini, Grok and Perplexity rather than one of them.

Can I pay to be cited?

Not directly in the engines. You can earn presence in cited third-party sources; some of those platforms have commercial listings, which is a business decision - but fabricated placements and fake reviews get filtered and are radioactive for trust.

How stable are citations?

They drift as models, indexes and sources change - a reason to remeasure the same questions on a schedule instead of trusting a one-time snapshot.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

Reporting AI visibility to your board: metrics that mean something

EntityRise — Tue, 13 Oct 2026 18:00:19 +0000

By Kamil Kępiński, founder of EntityRise

Four AI visibility metrics that survive board scrutiny - recommendation rate, description accuracy, citations, before/after delta - and what to refuse.

AI visibility earns a budget line only when it reports like one - and most of what gets reported today is noise: single screenshots, one blended score, unverifiable "share of AI" claims. Four metrics survive board scrutiny: recommendation rate on real buyer questions, description accuracy, citation presence in the sources AI engines use, and the before/after delta on identical questions. Each is measurable, repeatable and honest about its own confidence.

The four numbers worth reporting

Recommendation rate.Of your priority buyer questions, in what share of repeated runs across providers are you absent, mentioned or recommended - reported per funnel stage, because a strong shortlist rate with weak validation answers is a different problem than the reverse.

Description accuracy.When AI systems describe you, how often are the category, claims and facts right? Wrong descriptions convert worse than absence, and this number moves fastest after entity and markup fixes - an early proof point for the programme.

Citation presence.Of the sources AI engines actually cite for your category questions, how many mention you accurately? This is the leading indicator for live-search visibility.

Before/after delta.The same questions, same providers, remeasured after implementation. This is the only number that proves work happened - everything else is weather.

The numbers to refuse

Refuse one blended "AI score" with an opaque formula - it hides which system is failing and cannot survive a "what changed and why" question. Refuse single-run evidence - model answers vary between runs, so a screenshot proves availability of luck, not visibility. Refuse metrics without disclosed confidence - a serious report states how many runs completed, which provider tests failed and how much agreement supports each finding. If a vendor's reporting cannot answer these, the reporting is the product.

Reporting cadence and ownership

Quarterly reporting fits most B2B boards: long enough for implemented changes to reach live answers, short enough to catch drift - with monthly measurement underneath for the team running the work, because answers shift as models and sources change. Ownership belongs with whoever owns pipeline, not with a side-of-desk volunteer: the metrics above translate directly into shortlist presence, which translates into deals - which is the sentence that gets the budget approved.

Frequently asked questions

Can AI visibility be tied to revenue?

Directionally yes: shortlist presence on high-intent questions correlates with inbound quality, and validation-stage accuracy protects late-stage deals. Attribution is honest at the correlation level - anyone promising per-deal attribution from AI answers is overclaiming.

What is a good recommendation rate?

It depends on category competitiveness and question stage - which is why the meaningful benchmark is your own baseline, remeasured. Cross-company league tables mostly compare measurement setups, not companies.

How many questions and runs make a credible sample?

Enough repeated runs across several providers to see stability - as a reference point, an EntityRise audit records roughly 75 observations across ChatGPT, Claude, Gemini, Grok and Perplexity, with confidence labelled per finding.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

Write a category sentence machines can classify (with examples)

EntityRise — Tue, 13 Oct 2026 14:00:49 +0000

By Kamil Kępiński, founder of EntityRise

The one line that decides whether AI models can classify and recommend your company: the formula, good and bad examples, and where to deploy it.

A category sentence is one plain line stating what your company is, for whom, and in which category - the line ChatGPT, Claude, Gemini, Grok and Perplexity use to file you in their map of the market. Companies with a classifiable category sentence get recalled for category questions; companies with slogans in its place get skipped. It is the cheapest high-leverage fix in AI visibility: one sentence, deployed consistently, no engineering required.

The category sentence formula

The working pattern is: [Company] is a [precise category] that [core outcome] for [specific buyer]. Every slot matters. The category must be a term buyers and models actually use ("warehouse management software", "payroll provider for restaurants"), not an invented one. The outcome must be concrete enough to verify. The buyer must be narrow enough to match real questions - "for mid-size 3PLs" beats "for businesses of all sizes", because AI questions come with context and models match on it.

Examples: classifiable vs invisible

Invisible: "We empower organisations to unlock operational excellence through innovative solutions." No category, no buyer, nothing to file. Classifiable: "Acme is a warehouse management software platform that automates picking and inventory for mid-size 3PL operators." A model reading that line can answer "best WMS for 3PLs" with Acme in it. The test is mechanical: could a stranger - or a model - place you on a market map from the sentence alone, and name two competitors you should be compared with? If not, rewrite. One honest caveat: the sentence makes you classifiable, not automatically recommended - evidence and citations still decide the ranking within the category.

Where to deploy your category sentence

Consistency is the multiplier: the same sentence (verbatim or near-verbatim) on the homepage hero or first paragraph, the About page, the LinkedIn company profile, every directory listing, press boilerplate and the Organization schema description field. Models cross-reference sources; five matching descriptions build confidence, five different ones destroy it. Rebrands and pivots make this urgent - old category language on forgotten profiles keeps teaching models an outdated map. How models actually read your website is the technical half of the same job.

Frequently asked questions

Is this just a positioning statement?

It is positioning written for machine classifiability first: real category terms, verifiable claims, no metaphors. Good positioning and a good category sentence usually converge - the discipline of the formula gets you there faster.

Can I have two category sentences?

If you genuinely serve two categories, yes - one per category page, each internally consistent. What hurts is variation within one category: three phrasings of the same business read as ambiguity.

How do I know it worked?

Measure before and after: run category questions across providers and record whether recognition and description accuracy moved. Language changes are cheap to make and cheap to verify - there is no excuse for not measuring.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

Model memory vs live AI search: why you appear in one and not the other

EntityRise — Mon, 12 Oct 2026 18:00:22 +0000

By Kamil Kępiński, founder of EntityRise

AI assistants answer from two systems: model memory and live AI search. Why you appear in one and not the other - and the different fix each needs.

AI assistants answer from two different systems, and companies routinely score well in one while being invisible in the other. Model memory is what an assistant recalls about your market from training, with no internet access; live AI search is what it composes after browsing the current web and citing sources. ChatGPT, Claude, Gemini, Grok and Perplexity all operate both modes, and because the two are built differently, they fail differently - which is why serious AI visibility measurement always reports them separately.

How model memory decides who gets recommended

Model memory is a compressed picture of the public web as it looked during training. When a buyer asks "best providers of X" with search off, the model answers from that picture: companies that were consistently described, frequently mentioned and clearly classified before the training cutoff get recalled; everyone else blurs into the background. Two consequences follow. Recall lags reality - a repositioning or product launch takes time to enter memory. And recall rewards consistency - a company described the same way across hundreds of pages beats one with five conflicting descriptions, regardless of which is the better vendor.

How live AI search decides who gets cited

Live AI search flips the logic: the assistant runs web queries, reads a handful of current sources and composes an answer with citations. Here the question is not "what does the model remember about you" but "do the sources the engine trusts for this query mention you today". A company absent from model memory can dominate live answers by being present in the right comparison articles, directories and review sites - and a famous company can vanish from live answers when the cited sources skip it. Live results also move fast: publish better evidence, earn a citation, and answers can change within weeks.

Why model memory and live AI search diverge

The divergence is diagnostic. Strong memory, weak live search usually means legacy fame with a thin current citation footprint. Weak memory, strong live search is the profile of newer companies doing citations right - memory catches up with training cycles. Weak in both means foundational problems: unclassifiable category language, unverifiable claims or a machine-unreadable website. Strong in both is the goal, and it is measurable. This is also why one blended "AI visibility score" misleads: it hides which system is failing, and the fixes are different. The methodology EntityRise uses reports both systems separately for that reason.

What to do about memory and about live search

For model memory: entity consistency everywhere (one category sentence, identical naming), broad accurate presence across profiles and press, patience across training cycles. For live AI search: a citation map of the sources engines actually use in your category, verifiable evidence those sources can reference, and technical accessibility so engines can read you directly. The work overlaps but is not identical - and only remeasurement shows which side moved.

Frequently asked questions

Which matters more for B2B?

Both reach buyers, but live AI search usually moves first and moves faster, so most programmes start there while building the consistency that memory rewards.

Can I tell which mode produced an answer?

Often yes: cited sources and browsing indicators signal live search; instant uncited answers usually come from memory. Measured audits record the mode explicitly instead of guessing.

Why do I get different answers on different days?

Model outputs vary between runs, and live search depends on which sources the engine fetched that day. Single answers are anecdotes; repeated runs are measurement.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

The six types of buyer questions AI answers about your category

EntityRise — Mon, 12 Oct 2026 14:00:38 +0000

By Kamil Kępiński, founder of EntityRise

AI assistants answer six types of buyer questions, from problem discovery to brand validation. What each looks like and what to publish for it.

B2B buyers ask AI assistants six types of questions on the way to a purchase: problem discovery, category discovery, comparison, high purchase intent, brand validation and location-specific discovery. Each stage has its own question shape, its own answer format - and its own way of leaving you out. Mapping your AI visibility across all six is how you find where revenue actually leaks; it is also exactly how a structured audit builds its question set.

Stages 1-2: problem and category discovery

Problem discovery sounds like "how do we reduce picking errors in our warehouse" - no category named, no vendors expected, but assistants often end these answers by naming approaches and example providers. Category discovery is "best warehouse management software" - the classic shortlist question where two to five names carry the answer. Losing stage 2 is the loudest failure; losing stage 1 is the quietest, because you never learn the buyer existed. What wins here: classifiable category language and educational content that connects the problem to your category.

Stages 3-4: comparison and purchase intent

Comparison is "X vs Y for a company like ours" - assistants build feature-by-feature verdicts from whatever comparable evidence exists, and vendors without published, structured facts get represented by third-party guesses. High purchase intent is "X pricing", "how to migrate to X", "X implementation time" - questions from buyers close to a decision. If your own site does not answer them in machine-readable form, the assistant answers from someone else's version of your facts. What wins here: comparison pages, transparent pricing signals and concrete implementation detail.

Stages 5-6: validation and location

Brand validation is the last-mile check: "is X legitimate", "X reviews", "who is behind X". Assistants synthesise reviews, press, founder information and complaint patterns - and thin or inconsistent public records read as risk. Location-specific discovery ("X providers in Poland", "near us") filters the category map by geography, where directory data and local citations decide presence. What wins here: consistent entity facts, named people, verifiable customers and accurate local listings.

Measuring all six question stages - not just your favourite

Most companies check stage 2 once, see a result they like or hate, and stop. But the stages fail independently: strong shortlist presence with weak validation answers still kills deals late. A meaningful measurement runs real questions from every stage, repeatedly, across ChatGPT, Claude, Gemini, Grok and Perplexity, and reports each stage separately - the same structure EntityRise's methodology uses, with clients reviewing and editing the questions before the audit runs.

Frequently asked questions

Which stage should I fix first?

The one your measurement shows is leaking most against your deal flow. As a pattern, category discovery and validation move revenue fastest - but patterns are not your data.

Do buyers really ask assistants all six?

Not every buyer asks every stage, but across a buying committee the journey usually touches most of them - and each stage reached is a moment you are either present or absent.

Can I supply my own questions for an audit?

Yes - generated questions are reviewed and edited by the client before an EntityRise audit runs, and you can add the ones that matter most to your business.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

Schema markup for AI visibility: what actually gets read

EntityRise — Sun, 11 Oct 2026 18:00:16 +0000

By Kamil Kępiński, founder of EntityRise

The schema types AI systems actually use - Organization, Service, FAQPage, Article - what each does, what schema cannot do, and the mistakes that void it.

Schema markup is structured data embedded in your pages that states facts about your company outright - name, category, offerings, people, prices, questions and answers - instead of leaving machines to infer them from prose. For AI visibility its job is precise: schema removes ambiguity about entities. It does not make weak content strong, and it does not guarantee recommendations; it makes sure that what you say about yourself is read exactly as you meant it.

The schema types that matter for B2B AI visibility

Five types do most of the work across ChatGPT, Claude, Gemini, Grok and Perplexity. Organization - the anchor: legal name, category description (your category sentence belongs in it), logo, founders, sameAs links to LinkedIn and directories; this is the node everything else references. Service or Product - what you sell, in terms buyers use. FAQPage - your answer-first questions, marked up so engines can lift them cleanly. Article - authorship, dates and topic for every guide you publish; dateModified is your freshness signal. BreadcrumbList - cheap structural clarity. Everything beyond these is situational; nothing on this list is optional for a company that wants to be recommended.

What schema cannot do

Schema states claims; it does not prove them. A markup block declaring you "the leading provider" changes nothing, because models weigh verifiable evidence over self-description - and unverifiable superlatives get discounted with or without markup. Schema also cannot rescue unreadable pages: if the content is rendered client-side only or hidden from AI crawlers, the markup often goes unread with it (see how AI reads your website). Treat schema as the labelling layer on top of substance: essential, cheap, and worthless alone.

The schema mistakes that void it

Four failures repeat in audits. Markup that contradicts the visible page - different name, different description - which teaches machines to distrust both. Orphan nodes: Product and Article blocks that never reference the Organization @id, so nothing connects to your entity. Stale facts: old addresses, renamed products, departed founders still declared as current. And validation errors that silently break parsing - always test in a schema validator after every deployment, not once at launch. Consistency between markup, visible text and external profiles is the whole game.

Frequently asked questions

Does schema directly improve AI recommendations?

It improves recognition and description accuracy, which recommendations depend on. Measured audits regularly show description errors clearing after markup and consistency fixes - the recommendation lift follows evidence and citations.

JSON-LD or microdata?

JSON-LD: one block, easy to maintain, the format major systems parse most reliably.

Do AI models even read schema?

AI search engines and crawlers consume structured data in retrieval and indexing, and consistent markup feeds the public record models train on. The precise weighting per system is not public - another reason to measure your own before/after instead of trusting anyone's promise.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

How to get your company recommended by ChatGPT in 2026

EntityRise — Sun, 11 Oct 2026 15:02:16 +0000

By Kamil Kępiński, founder of EntityRise

A 2026 playbook for getting recommended by ChatGPT and other assistants: measure first, fix entity clarity, publish evidence, earn citations, remeasure.

Getting recommended by ChatGPT in 2026 comes down to five steps: measure where you stand, make your company machine-classifiable, publish evidence that third parties can verify, earn presence in the sources AI search engines cite, and remeasure the same questions to prove the change. The same playbook improves AI visibility in Claude, Gemini, Grok and Perplexity, because all of them reward the same thing: companies that are easy to recognise, classify and verify.

What changed for 2026

Two shifts define this year. AI search modes - assistants that browse before answering - now sit in front of a large share of B2B research, which means your live web footprint and citations matter as much as what models memorised in training. And buyers have learned to ask better questions: less "what is X" and more "X vs Y for companies like ours". Both shifts favour companies with concrete, comparable, verifiable public information - and punish vague positioning. For the wider discipline behind this playbook, see the GEO and AEO guide.

Step 1: measure before you touch anything

Run the questions your buyers ask - across providers, repeatedly - and record whether you are absent, mentioned or recommended, how you are described, who appears instead and which sources are cited. This baseline decides everything after it: which fixes matter, and whether they worked. The full measurement process EntityRise uses is public on the methodology page.

Steps 2-4: clarity, evidence, citations

Step 2 - entity clarity. One category sentence a machine can classify, used identically on your homepage, About page, LinkedIn and directories. Add Organization schema, server-render key content, publish llms.txt (see how AI reads your website).

Step 3 - verifiable evidence. Named customers, concrete outcomes with numbers, case studies with dates. Replace unverifiable superlatives with checkable facts - models discount what they cannot corroborate.

Step 4 - earned citations. Identify which sources AI engines cite for your category questions - review sites, comparison articles, industry databases - and earn accurate presence in them. This is the lever most companies skip, and the one that most often flips live-search answers.

Step 5: remeasure, and what to avoid

Remeasure the same questions on the same providers after implementation; the before/after delta is the only honest scoreboard. And avoid the shortcuts that backfire: fake reviews and fabricated testimonials (filtered, and reputationally radioactive), guarantee-sellers (no one controls independent AI providers), and one-off screenshot audits that mistake a lucky run for a trend.

Frequently asked questions

How long until results show?

Live AI-search answers can shift within weeks of publishing better evidence and earning citations. Model-memory recall follows provider training cycles - months, not days. Measure both separately and you always know which is moving.

Do I need different tactics for each assistant?

The foundations are shared. Providers differ in which sources they cite and how they search, which is why measurement runs across ChatGPT, Claude, Gemini, Grok and Perplexity rather than one of them.

Can I do this in-house?

The steps are public and doable in-house with time and discipline. What a specialist adds is the measurement infrastructure, the citation map for your category, and accountability for before/after proof.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

AI visibility in 2026: what changed and what to do about it

EntityRise — Sat, 10 Oct 2026 18:00:57 +0000

By Kamil Kępiński, founder of EntityRise

Four shifts define AI visibility in 2026: search-on by default, sharper buyer questions, professionalised citations and higher proof standards.

AI visibility in 2026 is shaped by four shifts: AI assistants now search the web by default for commercial questions, buyers ask sharper comparison questions, the citation layer has professionalised, and buyers of visibility services expect measured proof instead of dashboards. Companies that adapted to last year's rules - static content, one-time audits, blended scores - are quietly losing shortlist presence to companies working the current ones.

Shift 1: search-on is the default for commercial questions

Assistants increasingly browse before answering anything with buying intent, which moves the battleground from model memory toward the live web. The behavioural shift underneath is documented: 71% of B2B software buyers now rely on AI chatbots for software research, up from roughly 60% seven months earlier, and just 3% say assistants have not changed their research habits (G2, March 2026 survey of 1,076 buyers). Practical consequence: your current citation footprint - review platforms, comparison articles, directories - now shapes answers within weeks, not training cycles. Companies still optimising only their own website are optimising the smaller half of the system. For the discipline framing, see the GEO and AEO guide.

Shift 2: buyers ask context-rich questions

"Best CRM" is becoming "best CRM for a 40-person logistics company on a mid budget". Context-rich questions favour vendors with specific, verifiable, segment-level content - and punish generic positioning, because the model matches on the context. The winning content pattern of 2026 is narrow and concrete: named segments, real constraints, honest fit-and-misfit statements.

Shift 3: the citation layer professionalised

Which sources AI engines cite is no longer folklore - it is measurable per question, and categories now have identifiable citation maps. That cuts both ways: earning accurate presence in the right sources moves answers faster than ever, and neglected or wrong listings do damage at the same speed. Citation work has become the highest-leverage external activity in most B2B programmes.

Shift 4: proof standards went up

The market learned the difference between monitoring and improvement. Dashboards that show the problem weekly are now table stakes; the premium question is "who implements the fixes and proves the delta". Expect procurement to ask for before/after methodology, disclosed confidence and remeasurement on identical questions - and treat any vendor guaranteeing rankings as a red flag, because no one controls independent AI providers.

Your AI visibility plan for this quarter

Four moves, in order: measure your baseline on real buyer questions across ChatGPT, Claude, Gemini, Grok and Perplexity; fix entity clarity and machine readability (fast, cheap, compounding); build or refresh the citation map and earn the top gaps; publish two or three context-rich pages for your highest-value segments. Then remeasure the same questions - the delta is your 2026 scoreboard.

Frequently asked questions

Is model memory irrelevant now?

No - it still decides search-off answers and colours how models frame you even when browsing. It moves slower, which is exactly why the consistency work should start now.

Do these shifts apply outside tech?

Yes - manufacturing, logistics and professional services show the same patterns, often with weaker competition, which makes 2026 an outsized opportunity in traditional B2B.

What is the biggest mistake companies are making this year?

Buying visibility of the problem (another dashboard) instead of removal of the problem - monitoring without implementation documents decline in high resolution.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

AI search optimization checklist for B2B websites (2026)

EntityRise — Sat, 10 Oct 2026 14:00:52 +0000

By Kamil Kępiński, founder of EntityRise

A 20-point AI search optimization checklist for B2B websites: accessibility, entity clarity, buyer-intent content, evidence and citations, in order.

This AI search optimization checklist covers the twenty items that decide whether AI assistants and AI search engines - ChatGPT, Claude, Gemini, Grok and Perplexity among them - can find, understand, verify and recommend a B2B company. Work through the four groups in order - accessibility first, then clarity, then evidence, then citations - because each group depends on the one before it.

Technical accessibility (items 1-6)

AI systems can only recommend what they can read. See also how AI reads your website.

Key content server-rendered, not JavaScript-only.

AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) allowed in robots.txt - blocking them removes you from AI search.

Organization schema with name, category, logo and sameAs links.

llms.txt published, listing your most important pages.

Clean HTML headings that match visible text.

Critical claims in text, never only in images or PDFs.

Entity clarity (items 7-11)

Models recommend companies they can classify. Start with a category sentence a machine can parse.

One category sentence on the homepage a machine can parse.

The same sentence on About, LinkedIn and every directory profile.

Consistent legal and brand naming everywhere.

Founders and key people named and linked.

No contradictions between your site and external profiles - inconsistency reads as unreliability.

Buyer-intent content and evidence (items 12-16)

AI answers are built from the questions buyers ask.

Pages answering comparison questions ("X vs Y") in your category.

Use-case pages phrased the way buyers phrase them.

Named customers and concrete numbers instead of anonymous superlatives.

Case studies with dates and verifiable outcomes.

A public methodology or process page - the single most citable asset a service company can own.

EntityRise publishes its own measurement methodology for the same reason.

External citations (items 17-20)

Live AI answers are composed from cited sources. See AI citations explained.

Accurate presence on the review and comparison sites AI engines cite for your category.

Industry directory listings with consistent data.

Mentions in third-party articles buyers' questions surface.

A monitoring routine - or a monitoring service - that re-checks the same buyer questions on a schedule, because answers drift as models and sources change.

Frequently asked questions

How many items do most B2B sites fail?

In audit practice, most companies fail a large share of the accessibility and evidence groups - not because the fixes are hard, but because nobody had measured, so nobody owned the problem.

Which items matter most?

Sequence beats cherry-picking: accessibility failures silently cancel everything else, so fix 1-6 first, then clarity, then evidence, then citations.

Is this the same as an SEO checklist?

They overlap on technical hygiene, but items like llms.txt, AI-crawler access, category classifiability and citation mapping are specific to AI visibility. The differences are explained in our GEO vs AEO guide.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.

A realistic GEO strategy for 2026

EntityRise — Fri, 09 Oct 2026 20:40:36 +0000

By Kamil Kępiński, founder of EntityRise

What GEO can and cannot do in 2026, five workstreams in priority order, and how to prove any of it worked - without the hype.

A realistic GEO strategy for 2026 - covering ChatGPT, Claude, Gemini, Grok and Perplexity - accepts two facts most GEO content avoids: nobody controls what generative engines answer, and most GEO work is verifiable craft, not secret technique. Generative engine optimization - making your company more likely to be mentioned, described accurately and cited in AI answers - works when it is run like an engineering programme: baseline, prioritised workstreams, remeasurement. What GEO and AEO mean and how they differ is covered in our full guide; this is the operating plan.

What GEO can and cannot deliver

Can: measurable gains in recognition, description accuracy, citation presence and shortlist frequency on defined question sets - the components recommendations are built from. Cannot: guaranteed placement, permanent positions or immunity to model changes; answers drift as models, indexes and sources move. Any strategy - or vendor - promising the second list is selling weather control. Plan for the first list and for drift management, and 2026 becomes very winnable.

The five workstreams, in priority order

Baseline measurementReal buyer questions, repeated across providers, findings recorded with confidence. Everything else is sequenced by what this shows. The measurement process is public on the methodology page.

Machine readabilityServer-rendered content, AI crawlers allowed, schema, llms.txt. Cheap, fast, and failures here silently cancel the rest.

Entity clarityOne classifiable category sentence deployed identically across site and profiles.

Evidence and answerable contentVerifiable customers and numbers, plus pages that directly answer comparison, fit and validation questions.

Citation acquisitionMap the sources engines cite for your questions, earn accurate presence in the gaps. Highest external leverage, and the workstream most programmes never reach. How engines pick those sources is covered in AI citations explained.

This five-workstream sequence is exactly what an EntityRise Sprint implements - measured on both ends â†’

Cadence and drift management

Run it as loops, not a project: implement, remeasure the identical question set, catch drift, iterate. Quarterly loops fit most B2B categories; monthly measurement underneath catches source and model changes early. Budget honesty: workstreams 2-3 are days of work, 4-5 are ongoing capacity - which is why the realistic in-house question is not "can we" but "who owns loop five in March when the launch happens".

Frequently asked questions

Is GEO different from AEO in practice?

The labels differ more than the work does - both reduce to being recognisable, classifiable, verifiable and citable. If two invoices claim otherwise, ask what work is actually different; the full breakdown is in our GEO vs AEO guide.

How long until GEO shows results?

Readability and clarity fixes can move live answers in weeks; citation gains follow source-by-source; memory recall follows training cycles. An honest plan reports each lane separately.

Can this be done without a specialist?

Yes, with capacity and discipline - the workstreams are public. Specialists add the measurement instrument, the citation map and accountability for the delta.

About the author. Kamil Kępiński is the founder of Smophy Labs Inc., the company behind SmophyAI and EntityRise. Follow Kamil on X and LinkedIn.