DEV Community: EntityRise

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.

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.

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.

Does Google Penalize AI-Written Content? What the September 2026 Spam Update Really Means

EntityRise — Fri, 09 Oct 2026 19:56:00 +0000

By Kamil Kępiński, founder of EntityRise

The September 2026 spam update did not change Google's rules on AI content. What Google penalizes, what to ignore and why AI visibility can't be guaranteed.

Google does not penalize content because AI helped write it. It penalizes content that exists mainly to manipulate rankings or AI answers and gives readers nothing new, regardless of how that content was produced. The September 2026 spam update, which started on September 24, did not change this rule. The announcement of the update does not name any target. Google did update its guidance on AI-generated content on October 1, 2026, but the change asks for fact-checking, not a penalty for using AI.

Still, I believe the risk for companies that publish automated content is higher now than it was a year ago. Below I explain why, using what Google itself has published and what I see in our work at EntityRise.

Key points

On September 24, 2026, at 09:15 Pacific Time, Google started its fourth spam update of 2026. It applies globally and to all languages, and the rollout may take up to two weeks.
Google did not say what the update targets. Articles claiming it "targets AI content" are interpretations, not Google statements.
Since May 15, 2026, Google's spam policies explicitly cover attempts to manipulate generative AI responses in Search.
Google lists using generative AI tools to create many pages without adding value as an example of scaled content abuse.
Google says its Search ignores llms.txt files and does not need content split into small chunks for AI.
Nobody can guarantee a place in AI answers. Google says the same about crawling, indexing and serving pages.

What did Google change on September 24, 2026?

On September 24, 2026, Google released the September 2026 spam update, and it was the only ranking change Google announced that day. The Google Search Status Dashboard says the update applies globally and to all languages and that the rollout may take up to two weeks. It says nothing about AI content, links or any specific tactic.

Two other things happened the same day, and neither concerns AI-written content. Search Console began reporting searches made with images, such as Google Lens and Circle to Search, including in its report for generative AI features. Google's documentation changelog for that day has a single entry, about video structured data. Google's spam policies page was last updated on August 28, 2026, when Google adjusted how it enforces the site reputation policy in the European Economic Area.

What caught my attention is the length of the rollout window. Every earlier spam update in 2026 finished in less than three days. This one was announced with a window of up to two weeks. Google has not explained why, so I will not guess.

Google ranking updates in 2026, per the Search Status Dashboard (as of September 26, 2026)

Update	Started	Duration
Discover update	February 5	21 days, 17 hours
Spam update	March 24	19 hours, 30 minutes
Core update	March 27	12 days, 4 hours
Core update	May 21	11 days, 21 hours
Spam update	June 24	2 days, 1 hour
Spam update	August 18	2 days, 16 hours
Spam update	September 24	in progress, up to 2 weeks

Update, October 1, 2026: a week after the spam update started, Google updated its guidance on using generative AI content. The new text explains that generative models predict a likely sequence of words instead of retrieving facts, so their output can contain inaccuracies, and asks site owners to manually fact-check and review all AI-generated content before publishing, including titles, meta descriptions, structured data and image alt text. Google says the change brings its documentation in line with the Search Quality Raters guidelines.

Does Google penalize AI-written content?

No. Google judges content by whether it helps people, not by the tool used to write it. Its policy on scaled content abuse covers large amounts of unoriginal content that adds little or no value, regardless of how that content was created.

Google's examples of scaled content abuse include:

using generative AI tools or similar tools to generate many pages without adding value for users;

scraping feeds or search results and transforming them automatically, for example by swapping synonyms or translating;

stitching together content from different pages without adding value;

creating multiple sites to hide how much content is being produced.

Google's guidance also says that generative AI can be useful for researching a topic and adding structure to original content. The line does not run between AI and humans. It runs between value and volume.

One detail matters for online stores: Google Merchant Center requires AI-generated product images to carry metadata that marks them as AI-generated, and AI-generated product titles and descriptions to be submitted separately and labeled.

Why is automated blog content a bigger risk now?

This is where I see the real danger. Many products today sell automated blog writing as a way to grow in SEO. They produce a lot of articles quickly, and each of them tends to summarize what is already online.

That is exactly what Google calls commodity content in its guide to generative AI features in Search, published in May 2026. Google contrasts it with content that brings a unique point of view or first-hand experience, and it warns against recycling what others have already said or what a generative AI model could easily produce. The same guide says that creating separate pages for every variation of a query, including the related queries AI systems run in the background, violates the scaled content abuse policy when the goal is to manipulate rankings or AI answers.

Since May 15, 2026, Google's spam policies have also stated openly that trying to manipulate generative AI responses in Search is spam. The AI answer is no longer a side channel with softer rules.

So in my view, a blog filled with automated articles can lower a company's position in search and in AI answers instead of raising it. I am not saying AI has no place in writing. I am saying that volume without a person who knows the subject has become a liability.

What does Google tell you to ignore?

Google's guide lists several popular "AEO" and "GEO" tactics that Google Search does not need:

llms.txt files and other special AI files or markup; Google says they neither help nor harm rankings;

splitting content into small "chunks" for AI;

rewriting content just for AI systems;

chasing inauthentic mentions around the web;

overfocusing on structured data, which helps with rich results but is not required for AI features.

We keep an llms.txt file on entityrise.ai, because other assistants may read it. For Google Search it is neutral, and Google says so openly.

Google also writes that, from its perspective, optimizing for generative AI search is still SEO. For Google Search, I agree. But Google speaks only for Google. ChatGPT, Claude, Grok and Perplexity are separate systems with their own models and sources, and Google's guide does not describe how they choose what to recommend. For how those systems differ from classic SEO, see AI visibility vs SEO.

Why can nobody guarantee visibility in AI answers?

Nobody can guarantee that AI assistants will recommend your brand, because the systems that decide keep changing. Google itself writes that a page meeting every requirement is still not guaranteed to be crawled, indexed or shown. About outside tools it is even more direct: "No third-party tool has access to our internal ranking or AI systems."

Google also explains that its AI features are built on its core ranking and quality systems: they retrieve pages from the Search index and run related queries in the background to gather more results. In practice, every core or spam update can change which pages AI Overviews and AI Mode rely on. In 2026 alone, Google confirmed seven ranking updates in less than eight months, between February 5 and September 24.

This matches what I see in my work. What helps AI show your company today can stop working in two or three months, when Google changes its systems or a model gets a new version and starts giving more weight to other sources of truth about your company. That is why every EntityRise audit records the model version of each AI provider and the time of the test. A result is true for a specific moment, not forever. The process is public on our methodology page.

So if someone promises you 100% visibility in AI, permanently, they are not telling you the truth. They are monetizing your uncertainty. What you can do, and in my opinion should do, is stay fully compliant with the rules, measure regularly and keep working on the signals AI systems use to understand your company.

Is search moving from Google to AI?

Search is moving to AI, and a large part of that AI is Google itself. According to figures Google has published:

AI Overviews have over 2.5 billion monthly active users (Sundar Pichai, Google I/O, May 19, 2026);

AI Mode passed 1 billion monthly active users (Google I/O, May 19, 2026, repeated on Alphabet's earnings call on July 22, 2026);

Google says it sends billions of clicks to websites every week through AI features in Search (July 22, 2026);

since August 31, 2026, every site in Search Console has a dedicated report on its impressions in Google's generative AI features.

People increasingly ask for a recommendation and get a short answer instead of ten links. That is where many buying decisions start today. And because the models behind those answers keep evolving, the way they perceive your company and your products keeps changing too. That is the core of AI visibility.

Where EntityRise stands: a person behind every change

This is the part I care about most. At EntityRise, measurement is automated, because it has to be repeatable: you approve 20 buyer questions, we seal the set before the first test and run it across ChatGPT, Claude, Gemini, Grok and Perplexity, which gives up to 100 observations. Everything that changes a client's website, content or presence elsewhere on the web is done by people, and each change is verified live. In the Expert Audit, a person checks every finding before it reaches the client.

Our team has been working together since 2017, across AI, SEO, GEO and AEO. We have watched these systems change for years, and that is exactly why we do not promise permanence. We tell clients how it really works: visibility can be measured and improved, but it has to be maintained.

What I would do now

Stop publishing in bulk, especially pages created for every keyword variant.

Make sure every article contains something only you can say: your own data, a case from practice, a decision or a mistake and what it taught you.

Sign articles with the name of a real person and link to an author page.

Check your impressions in the Generative AI performance report in Google Search Console.

Measure how ChatGPT, Claude, Gemini, Grok and Perplexity describe you and your competitors, and repeat the measurement after major updates.

Keep the technical basics in order: pages indexable, crawlable and allowed to show a snippet.

Treat any promise of guaranteed AI visibility as a warning sign.

Frequently asked questions

Does Google penalize content written with AI?

No. Google evaluates whether content helps users, regardless of how it was created. Using generative AI tools to produce many pages without adding value is treated as scaled content abuse under Google's spam policies.

What did the September 2026 spam update target?

Google did not say. The Search Status Dashboard states only that the update started on September 24, 2026, applies globally and to all languages, and may take up to two weeks.

Does llms.txt help visibility in Google Search?

No. Google says its Search ignores llms.txt files, so they neither help nor harm rankings. They can still be useful for other AI systems.

Can anyone guarantee that AI assistants will recommend my company?

No. Google states that meeting every requirement does not guarantee crawling, indexing or serving, and AI models change their sources and versions over time. Visibility in AI can be measured and improved, but not guaranteed.

How can I check my visibility in Google's AI features and other assistants?

Google Search Console has a Generative AI performance report, available to all sites since August 31, 2026. For ChatGPT, Claude, Gemini, Grok and Perplexity, EntityRise measures how each assistant describes and recommends your company.

Sources

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 tools in 2026: monitoring dashboards vs done-for-you services

EntityRise — Fri, 09 Oct 2026 19:25:44 +0000

By Kamil Kępiński, founder of EntityRise

The 2026 market in three buckets - monitoring dashboards, agencies, measure-and-implement services. What each gives you and how to choose.

The AI visibility market in 2026 sorts into three buckets: monitoring dashboards that show how ChatGPT, Claude, Gemini, Grok and Perplexity mention you, marketing agencies that produce content and PR, and measure-and-implement services that do both ends - baseline measurement, the fixes, and remeasurement. The buckets are complementary on paper and confused in practice, which is how companies end up paying for visibility of a problem nobody is removing.

Bucket 1: monitoring dashboards

What they give you: recurring tracking of prompts across assistants, share-of-voice charts, alerting when answers change - genuinely useful telemetry, especially for teams with in-house capacity to act on it. What they cannot give you: the work. A dashboard identifies that ChatGPT recommends your competitor; someone still has to fix the category language, publish the evidence and earn the citations. Budget rule of thumb: a dashboard is worth it when a named person owns acting on what it shows - otherwise it is a subscription to watching. Which numbers are worth tracking at all is covered in AI visibility metrics.

Bucket 2: marketing agencies

What they give you: production capacity - content, PR, sometimes technical SEO - and increasingly a "GEO" line on the proposal. What too many cannot give you: measurement. If the engagement does not begin with a recorded baseline on real buyer questions and end with remeasurement of the same questions, results stay unprovable in both directions - you cannot see success and you cannot catch failure. The vetting question is one sentence: "show me the before/after instrument".

Bucket 3: measure-and-implement services

The combination bucket: baseline measurement across providers, implementation of the fixes the measurement prioritises, then remeasurement on identical questions - one accountable team for the whole loop. This is the model EntityRise runs (the measurement side is public on the methodology page), and the honest trade-off is capacity: doing both ends properly limits how many clients a team takes per quarter, which is why this bucket suits companies that want the problem removed more than they want another login.

How to choose an AI visibility tool by situation

Strong in-house web and content team with free capacity: dashboard plus your own execution can work - add an independent baseline so progress is provable. Existing agency you trust: keep it, and add measurement around its work so the GEO line item earns its keep. No spare capacity and revenue leaking from AI shortlists now: the measure-and-implement bucket exists precisely for you. In every scenario the constant is the same: measurement first, identical-question remeasurement after - tools and vendors are interchangeable, proof is not.

Frequently asked questions

Should I buy a dashboard and a service?

Sometimes - continuous monitoring between implementation cycles is how drift gets caught. Measure-and-implement engagements often include monitoring, so check before paying twice.

Why are specific tools not named and ranked here?

Because the market shifts monthly and a static ranking would mislead within a quarter. The bucket logic is stable; apply it to whatever shortlist is current when you read this.

What does AI visibility work cost in 2026?

Dashboards run from low monthly subscriptions; agency and service engagements vary with scope. EntityRise publishes its pricing openly - see the pricing page for current numbers.

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 statistics 2026: how B2B buying research moved into assistants

EntityRise — Fri, 09 Oct 2026 15:15:44 +0000

By Kamil Kępiński, founder of EntityRise

Sourced AI search statistics for 2026: adoption among B2B buyers, how AI changes vendor selection, and what it does to pipelines. Every figure dated.

Reviewed quarterly.

This page collects sourced statistics on how AI assistants changed B2B buying research in 2026. Every number carries its original source, sample and publication date, because AI visibility statistics age fast and circulate loosely - and a statistics page that cannot show its sources is marketing wearing a lab coat. The figures are grouped into adoption, buying behaviour and business impact, and the page is reviewed quarterly.

How to read AI search statistics

Three habits keep you honest. Check the denominator: "x% use AI" means something different for consumers, knowledge workers and B2B buying committees, and most headline numbers come from software buyers specifically. Check the date: adoption curves this steep make 18-month-old figures historical artefacts - G2 measured a jump from roughly 60% to 71% in seven months. Check the incentive: vendors publish statistics that flatter their category, EntityRise included, which is why every figure below names its source and sample so you can weigh it yourself.

Adoption: who is asking assistants

51% of B2B software buyers now start their research with an AI chatbot more often than with Google (G2, The Answer Economy, survey of 1,076 B2B software buyers and decision-makers, March 2026, published April 2026).

71% of B2B software buyers rely on AI chatbots for software research, up from roughly 60% seven months earlier (G2, same survey).

86% of B2B buyers increased their use of AI chatbots for software research over the past year, and just 3% say AI chatbots have not meaningfully changed their research habits (G2, same survey).

66% of US business professionals regularly use AI specifically to research products, vendors or solutions for work (Semrush, survey of 600+ US business professionals, published 2026).

94% of business buyers report using AI during their buying process, up from 89% a year earlier (Forrester, Buyers' Journey Survey of nearly 18,000 global business buyers, reported in The State Of Business Buying, 2026, January 2026).

Provider spread for product research: ChatGPT 71%, Gemini 61%, Copilot 45%, Perplexity 18%, Claude 14% (Semrush, same survey) - buyers move between several assistants, which is why single-provider checks miss most of the picture.

Buying behaviour: what AI changes in vendor selection

69% of buyers chose a different software vendor than they originally planned, based on AI chatbot guidance (G2, March 2026).

33% - one in three - purchased from a vendor they had never heard of before it appeared in an AI answer (G2, March 2026).

92% of B2B buyers say AI has shaped their vendor shortlist, and 83% say it influenced their final vendor decision (Semrush, 2026).

41% say comparing vendor strengths and weaknesses is their top use case for AI chatbots in software research, ahead of basic product research and vendor identification (G2, March 2026).

82% of B2B software buyers sourced software recommendations from an AI chatbot in the last two years (G2, 2026 Buyer Behavior Report, survey of more than 1,000 buyers, July 2026).

Fragmentation: the same question, different AI answers

12% of cited sources match across ChatGPT, Perplexity and Google AI answers (Passionfruit, 2026; the 15,000-query sample as reported in industry coverage).

11% of domains are cited by both ChatGPT and Perplexity, in a study of 680 million citations (Averi, B2B SaaS Citation Benchmarks Report, March 2026).

More than 60% of 1,600 test queries got incorrect citation answers across eight AI search engines, with Perplexity lowest at 37% (Tow Center for Digital Journalism, Columbia University, 200 news articles tested on each engine, March 2025).

Taken together, these are the strongest published argument against single-provider checks: the engines disagree about which sources to trust, so visibility measured in one of them says little about the others. How engines choose those sources is covered in AI citations explained, and the measurement protocol that accounts for it is on the methodology page.

Business impact: what it does to pipelines

1.08% of total website traffic came from AI referrals across 13,770 domains studied - AI referral volume is still small (Conductor, 2026 AEO / GEO Benchmarks Report).

14.2% conversion rate for AI-referred visitors versus 2.8% for Google organic, roughly a 5x difference, across 312 B2B technology firms' GA4 and CRM data (Opollo, 2026 AI Search Benchmark Report).

0.5% of sessions from AI search produced 12.1% of signups for one B2B SaaS company - a 23x differential, and a single-company ceiling rather than a benchmark (Ahrefs, published June 2025).

31% better conversion from AI referrals than non-AI traffic during the 2025 holiday season, across more than a trillion visits to US retail sites (Adobe Digital Insights, January 2026).

What the pattern across these sources supports, stated conservatively: AI referrals are low in volume and high in intent, and volume-based reporting systematically undervalues the channel. The honest reading is that AI visibility is not yet a traffic story - it is a shortlist story, and the shortlist forms before any traffic exists to measure. For what changed around these numbers this year, see AI visibility in 2026.

Frequently asked questions

Can I cite these statistics?

Yes - cite the original source named beside each figure, with its publication date, rather than this page. That is also how we would want anyone treating EntityRise measurement data.

Why do published AI statistics contradict each other?

Different denominators, samples, dates and definitions of "AI search". Contradiction between honest studies is normal; a vendor citing only the flattering half is the thing to watch for.

Which figures here are strongest?

The ones with named samples and dates: G2's 1,076-buyer survey, Semrush's 600+ professional survey, Adobe's trillion-visit dataset and Columbia's Tow Center citation study. Details marked "as reported in industry coverage" come from research summarised by third parties - directionally useful, worth verifying at source before quoting in a board deck.

What is the one number that matters for my company?

None on this page - the number that matters is how often assistants recommend you on your buyers' questions, and that is measurable directly.

Sources

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