The complete guide to AI search visibility

This guide is for B2B marketing leads, SEO owners and founders who keep hearing that buyers now “ask ChatGPT” before they ever reach a vendor website, and who need to decide what, if anything, to change. It covers ChatGPT search, Perplexity, Google AI Overviews and AI Mode, Claude and Microsoft Copilot.

After reading it you should be able to check whether these systems can reach your site at all, separate what the platforms have documented from what vendors and consultants are guessing, pick a measurement setup that fits your budget, and run a 90-day plan without buying a new category of software on day one.

It is not for teams that have never done basic SEO. If your site is not properly indexed in Google and Bing, fix that first.

Summary

• The platforms’ own documentation says the entry ticket is ordinary: be crawlable, be indexed, be eligible for a snippet. There is no secret file or markup.
• Search crawlers and training crawlers are separate. You can block model training and still appear in AI search, but blocking the search bot removes you.
• The strongest public correlation with AI visibility is how often your brand is mentioned on other sites, not your backlink count.
• AI answers change from run to run, so “rank #1 in ChatGPT” is not a meaningful metric. Share of appearances across many runs is.
• llms.txt and special schema have weak or no documented effect on AI search inclusion today. Google says it ignores llms.txt for Search.
• First-party measurement now exists: Google Search Console and Bing Webmaster Tools both report AI-feature visibility, and ChatGPT tags its referral links.
• Most of the work is SEO and PR you should already be doing, aimed at the questions buyers ask during evaluation.

1. What GEO, AEO and “LLM visibility” actually mean

Three labels circulate for roughly the same goal.

GEO (generative engine optimization) comes from a 2023 academic paper that defined “generative engines” as systems that retrieve information from several sources and use a large language model to write a single answer. The authors proposed GEO as a way for content creators to improve how visible their content is inside those answers.¹1 Source: Aggarwal et al., “GEO: Generative Engine Optimization,” arXiv 2311.09735, accessed Oct 2026. Abstract reports visibility gains of up to 40% on its benchmark and notes efficacy varies by domain.

AEO (answer engine optimization) is an older industry term for the same idea, now used interchangeably with GEO.

LLM visibility or AI visibility is the measurement side: how often your brand is named, recommended or cited as a source when someone asks an AI system a question relevant to what you sell.

Two outcomes matter and they are different:

• Mention. The answer names your company (“Tools commonly used for this include A, B and your product”).
• Citation. The answer links to a page on your site, or to a third-party page about you, as a source.

A mention shapes the shortlist. A citation can send a click. Track both.

Three more terms. ICP (ideal customer profile) is the type of company you sell to. Robots.txt is the file at a domain’s root telling crawlers which paths they may fetch. Grounding is when an AI system retrieves live documents to base its answer on.

Where AI visibility sits among a company’s other revenue systems is covered in the complete guide to building a B2B revenue stack.

2. How AI search systems find and cite pages

Each system works in two layers: a language model trained on a fixed snapshot of text, and a retrieval layer that searches an index at question time and feeds documents to the model. Citations come from retrieval, the layer you can influence in weeks rather than years.

Google AI Overviews and AI Mode

Google is explicit that these features sit on top of its normal search index. To be eligible, a page must be indexed and eligible to be shown in Google Search with a snippet, and Google says there are no additional requirements to appear in AI Overviews or AI Mode.²2 Source: Google Search Central, “AI features and your website”, accessed Oct 2026. Google also describes a technique it calls query fan-out: both features may run multiple related searches across subtopics and data sources to assemble a response, which is why they can show a wider set of links than a classic results page.²

For B2B, fan-out matters. A question about automating SOC 2 evidence (SOC 2 is the security attestation many software buyers require) may trigger sub-searches on pricing, integrations and audit readiness. A page that answers one sub-question well can be cited without ranking for the full question.

Google’s newer guidance on optimizing for generative AI features adds one administrative requirement: your site must be included in the generative AI features setting in Search Console.³3 Source: Google Search Central, “Optimizing for generative AI features on Google Search”, accessed Oct 2026. Announced on the Search Central blog in May 2026. Check it.

The audience is large. On Alphabet’s July 2025 earnings call, CEO Sundar Pichai said AI Overviews had 2 billion monthly users.⁴4 Source: TechCrunch, “Google’s AI Overviews have 2B monthly users, AI Mode 100M in the US and India”, accessed Oct 2026. Figure stated by Sundar Pichai on the Q2 2025 earnings call.

ChatGPT search

OpenAI runs separate crawlers for separate jobs. OAI-SearchBot surfaces websites in ChatGPT’s search features. GPTBot crawls content that may be used to train OpenAI’s foundation models. ChatGPT-User handles certain actions a user triggers, and OpenAI says it is not used to crawl the web automatically.⁵5 Source: OpenAI, “Overview of OpenAI crawlers”, accessed Oct 2026. Sites that block OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links.⁵ OpenAI’s publisher FAQ says any public website can appear in ChatGPT search, provided OAI-SearchBot is not blocked.⁶6 Source: OpenAI Help Center, “Publishers and Developers FAQ”, accessed Oct 2026 via search result text; the page blocked direct retrieval.

Perplexity

Perplexity documents two agents. PerplexityBot surfaces and links websites in Perplexity’s results, and the company says it is not used to crawl content for foundation models. Perplexity-User fetches a page when a user asks a question, and because a person requested it, Perplexity says this fetcher generally ignores robots.txt.⁷7 Source: Perplexity, “Perplexity crawlers”, accessed Oct 2026.

Claude

Anthropic lists three bots. ClaudeBot collects content that may contribute to model training. Claude-User fetches pages when a person asks Claude a question. Claude-SearchBot indexes content to improve search result quality, and Anthropic says blocking it may reduce your visibility in Claude’s search answers. Anthropic says its bots honor standard robots.txt directives and that blocking by IP address may not reliably opt you out.⁸8 Source: Anthropic Help Center, “Does Anthropic crawl data from the web, and how can site owners block the crawler?”, accessed Oct 2026.

Microsoft Copilot and Bing

Copilot is grounded in Bing. Microsoft’s Bing team says grounding shares crawl and ranking infrastructure with classic search but looks for information with clear provenance, tracking factual fidelity, attribution, freshness and conflicts between sources.⁹9 Source: Microsoft Bing Blogs, “Evolving role of the index: From ranking pages to supporting answers”, accessed Oct 2026. An unverified Bing Webmaster Tools account means you are missing the first-party Copilot citation report (chapter 6).

The crawler map

Platform	Search / answer crawler	Training crawler	User-triggered fetcher	Key point
OpenAI (ChatGPT)	OAI-SearchBot	GPTBot	ChatGPT-User	Blocking OAI-SearchBot removes you from ChatGPT search answers5
Perplexity	PerplexityBot	None listed; PerplexityBot “not used” for model training	Perplexity-User	User fetcher generally ignores robots.txt7
Anthropic (Claude)	Claude-SearchBot	ClaudeBot	Claude-User	Each bot is controlled separately in robots.txt8
Google (AI Overviews, AI Mode)	Googlebot (normal Search crawl)	Google-Extended token for Gemini training and grounding	n/a	Google-Extended does not affect Search inclusion or ranking1010 Source: Google Crawling Infrastructure, “Google’s common crawlers”, accessed Oct 2026. Google-Extended is a robots.txt token with no separate user agent.
Microsoft (Copilot)	Bing index	n/a	n/a	Citations reported in Bing Webmaster Tools1111 Source: Bing Webmaster Blog, “Introducing AI Performance in Bing Webmaster Tools Public Preview”, accessed Oct 2026. Dated February 10, 2026.

The upshot: if someone was asked to “block AI bots” a couple of years ago, they may have blocked OAI-SearchBot or Claude-SearchBot along with the training bots. Check robots.txt and your CDN or firewall rules first.

The same crawler and tool-access questions come up when companies deploy agents of their own; see the complete guide to AI agents in enterprise operations.

3. What is known, and what is speculation, about getting cited

This is where most published advice goes wrong. It helps to sort claims by who is making them.

What the platforms have documented

Google has said the most, and it is mostly a list of things you do not need. Its AI features page says you don’t need new machine-readable files, AI text files or markup to appear, and there is no special schema.org structured data to add.² Its optimization guide goes further: there is no requirement to break content into tiny pieces for AI, structured data isn’t required for generative AI search, and llms.txt files will neither harm nor help visibility in Google Search.³

On content, Google’s advice is to publish unique, first-hand material rather than recycling what others have already said, and to write for people rather than to manipulate AI responses. It warns that mass-producing pages for query variations risks breaching its scaled content abuse policy.³ It also warns that seeking inauthentic mentions across the web “isn’t as helpful as it might seem,” and that no third-party tool has access to its internal ranking or AI systems.³

OpenAI, Perplexity and Anthropic document access and crawler behavior. None of them, in the pages we reviewed, publishes ranking or citation factors.⁵⁷⁸

What independent research suggests

The original GEO paper tested content changes on a benchmark of queries and reported that its methods could boost visibility in generative engine responses by up to 40%, while noting that how well the strategies work varies across domains.¹ “Up to” is a ceiling from a lab benchmark, not an expected result for your site.

Ahrefs analyzed 75,000 brands in May 2025 and correlated various factors with how often each brand appeared in Google AI Overviews. Branded web mentions had the strongest correlation (0.664), followed by branded anchor text (0.527) and branded search volume (0.392). Domain Rating (0.326), referring domains (0.295) and backlinks (0.218) were weaker.¹²12 Source: Ahrefs, “An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied)”, accessed Oct 2026. Spearman correlations; published May 2025 by a vendor of AI visibility tools. Ahrefs itself stresses that this is correlation, not causation, and that it measured one Google surface. Ahrefs also sells an AI visibility product, so weigh it accordingly.

What is speculation

Claims you will hear without primary support: that a word count, FAQ block or “answer in the first sentence” format causes citations, or that freshness dates alone move them. Some may be partly true. None is documented by the platforms. Treat them as tests to run.

4. The role of third-party mentions, reviews and earned media

If the Ahrefs correlation holds even roughly, much of your AI visibility is decided off your own site.¹² That fits how retrieval works. Ask about usage-based billing options for mid-market SaaS and the system pulls comparison articles, review sites, community threads and analyst pages. Your product page is one voice among many, and the least neutral.

Pew Research Center’s tracked-browsing study of 900 US adults found that Wikipedia, YouTube and Reddit were the most frequently linked sources in both AI summaries and standard Google results, together accounting for 15% of AI summary sources.¹³13 Source: Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results”, accessed Oct 2026. Browsing data from 900 US adults, March 2025. B2B categories have their own equivalents: review platforms, industry publications, integration marketplaces, partner directories and practitioner communities.

What this means in practice:

• Review sites. Keep profiles current, with accurate pricing model, integrations and ICP. Ask real customers for reviews after onboarding or renewal.
• Earned media. Original data that journalists and analysts can cite, such as an anonymized usage benchmark, tends to get repeated.
• Comparison content written by others. If you are absent from the articles comparing your category, answers built from them will omit you. Ask to be included where you genuinely fit.
• Communities. Employees answering questions in their own name is legitimate. Seeding fake recommendations is not, and Google warns inauthentic mentions are less useful than they look.³

Who this is not for: if you sell into a tiny niche where buyers know every vendor by name, AI visibility work will move little. Your money is better spent on direct relationships.

5. Structured data and llms.txt: an honest read of the evidence

Structured data

Structured data is machine-readable markup (usually JSON-LD using the schema.org vocabulary) that labels what a page contains. Google says it gives “explicit clues about the meaning of a page” and can enable rich results, while stating that eligibility is not guaranteed.¹⁴14 Source: Google Search Central, “Introduction to structured data markup in Google Search”, accessed Oct 2026. For AI features specifically, Google says structured data isn’t required and there is no special markup to add, though it remains useful for rich results.³

Our read: keep the structured data you would use for classic SEO, matched to visible text. No platform has documented that it drives AI citations.

llms.txt

llms.txt is a proposal, published by Jeremy Howard in September 2024, for a Markdown file at a site’s root that gives AI agents a curated overview and links to LLM-friendly versions of pages. Its authors say they expected it to be mainly useful at inference time, when an agent is helping a user, rather than for training.¹⁵15 Source: llmstxt.org, “The /llms.txt file”, accessed Oct 2026.

The evidence that it affects AI search visibility is thin. Google’s John Mueller wrote in June 2025 that no AI system currently uses llms.txt, based on what Google could see in server logs.¹⁶16 Source: Search Engine Roundtable, “Google: No AI System Currently Uses LLMs.txt”, accessed Oct 2026 via search result text. Reports a June 2025 Bluesky post by John Mueller. In 2026 Google’s own optimization guide said Search ignores the file and that it will neither harm nor help visibility in Google Search.³ None of the OpenAI, Perplexity or Anthropic crawler pages we reviewed mentions it.⁵⁷⁸

For a developer docs site that coding agents read, an llms.txt is cheap and may help them navigate. For a marketing site seeking citations, it is low priority.

6. Measuring AI visibility

Measurement is where budgets go first and where expectations need the most correction.

Why “rank in ChatGPT” is the wrong metric

SparkToro and Gumshoe had about 600 volunteers run 12 prompts through ChatGPT, Claude and Google’s AI results, 2,961 runs in total, in late 2025. For ChatGPT and Google’s AI, they estimated less than a 1-in-100 chance that two of 100 runs would return the same list of brands, and roughly 1 in 1,000 for the same list in the same order.¹⁷17 Source: SparkToro, “NEW Research: AIs are highly inconsistent when recommending brands or products”, accessed Oct 2026. Published January 28, 2026; not peer reviewed; co-author works at an AI tracking vendor. The authors concluded that position in an AI answer is not a useful metric, but that the percentage of runs in which a brand appears, measured across many prompts and many runs, is defensible. Gumshoe sells AI tracking, a conflict the study discloses.¹⁷

First-party data you already have access to

Google Search Console. Clicks and impressions from AI Overviews and AI Mode are included in the standard Performance report under the “Web” search type.² In June 2026 Google added dedicated Search Generative AI performance reports showing how often your URLs appeared in generative AI features, broken down by page, country, device and date. A note on the announcement says they reached all websites worldwide as of August 31, 2026. They report impressions, not queries.¹⁸18 Source: Google Search Central Blog, “Introducing Search Generative AI performance reports in Search Console”, accessed Oct 2026. Posted June 3, 2026.

Bing Webmaster Tools. The AI Performance report, in public preview since February 2026, shows total citations, average cited pages per day, page-level citation counts and a sample of “grounding queries” across Copilot, AI summaries in Bing and select partner integrations. Microsoft says citation counts do not reflect importance, ranking or placement within an answer.¹¹ A June 2026 update added intent and topic breakdowns, citation share and comparisons.¹⁹19 Source: Microsoft Bing Blogs, “New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare”, accessed Oct 2026 via search result text.

Web analytics referrals. OpenAI says ChatGPT adds utm_source=chatgpt.com to referral URLs, so those visits show up in tools like Google Analytics.⁶ For other assistants, segment on the referring domains you see in your own data, such as perplexity.ai or claude.ai. Expect undercounting: many buyers read the answer and later search your brand directly.

Server logs. Logs show whether the search crawlers are fetching your pages. It is the cheapest warning that something is blocked.

Expect fewer clicks per appearance

Pew found that Google users clicked a traditional result on 8% of visits to pages with an AI summary versus 15% without one, and clicked a link inside the summary on 1% of visits.¹³ AI visibility is partly a shortlist metric. Judge it alongside branded search, direct traffic and “how did you hear about us” answers.

Tool categories and pricing models

Third-party trackers run your list of prompts through the AI platforms on a schedule and report mention rate, citations and competitor share. Pricing is usually tied to prompts tracked, engines covered, and check frequency.

Category	Examples	How it is priced	List price as of October 2026	Best for
Free first-party reports	Google Search Console, Bing Webmaster Tools	Free	$0	Everyone; start here1811
AI add-ons inside SEO suites	Ahrefs Brand Radar, Semrush AI Visibility Toolkit	Monthly add-on to an existing subscription; prompt packs or check volumes	Ahrefs: Brand Radar AI “from $199/mo”; custom prompt packs from $50/mo for 2,500 checks.2020 Source: Ahrefs, “Pricing”, accessed Oct 2026. List prices; Brand Radar is an add-on. Semrush: AI Visibility Toolkit $99/mo per its knowledge base; bundled SEO + AI Search plans from $199/mo2121 Source: Semrush Knowledge Base, “AI Visibility Toolkit”, accessed Oct 2026. List price; no free trial per the page. 2222 Source: Semrush, “Plans and pricing”, accessed Oct 2026. Monthly billing list prices.	Teams already paying for the suite
Dedicated AI visibility platforms (self-serve)	Peec AI	Tiered monthly plans by tracked prompts, models and frequency	Plans listed (Starter, Pro, Advanced); prices not shown on the page we reviewed2323 Source: Peec AI, “Pricing”, accessed Oct 2026. The page names ChatGPT, Perplexity and Gemini as tracked engines.	Marketing teams wanting a focused tool
Enterprise AI visibility platforms	Profound	Custom annual contracts; free trial	Custom; trial of 50 prompts for 7 days on three engines2424 Source: Profound, “Pricing”, accessed Oct 2026.	Large brands tracking many products or regions
Manual sampling	Spreadsheet plus repeated prompts	Staff time	$0 in software	Small teams validating whether this matters

Check engine coverage before buying. Semrush’s prompt tracking covers Google AI Mode and ChatGPT, per its knowledge base;²¹ Profound’s enterprise tier lists nine engines including Claude, Copilot and Perplexity.²⁴ If your buyers live in Microsoft 365, a tool without Copilot misses them.

Who should not buy a tracker yet: a team that has not fixed crawler access or opened the free Search Console and Bing reports.

7. How AI search visibility relates to classic SEO

The relationship is closer than the GEO label implies.

• Indexing is shared. Google’s AI features use its normal index and ranking systems,³ and Copilot draws on Bing’s.⁹ Invisible in classic search usually means invisible here.
• Quality guidance is shared. Original, first-hand, people-first content.³
• Authority is weighted differently. Classic SEO leaned on links; the correlation data points more to unlinked brand mentions.¹²
• The unit of success differs. Appearance share across varied prompts replaces rank, and clicks per appearance are lower.¹³¹⁷
• Query shape differs. Prompts are longer and get fanned out into sub-queries,² so pages answering specific evaluation questions beat broad category pages.

Two areas need new work: the crawler review in chapter 2, and Bing, which many B2B teams ignored for a decade and which now feeds Copilot.¹¹

Search is one source of pipeline. For the outbound side, see the complete guide to outbound in 2026.

8. A 90-day plan

This assumes one marketer owns it, with a few hours a week from engineering and PR.

Days 1 to 30: access and baseline

1. Audit robots.txt and bot rules. Allow OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot on public pages, including at the CDN and firewall. Decide separately, with legal, about training crawlers (GPTBot, ClaudeBot, Google-Extended). OpenAI and Perplexity say changes take about 24 hours to reflect.⁵⁷
2. Verify Bing Webmaster Tools and open the AI Performance report.¹¹
3. Check Search Console, including the generative AI features setting and the generative AI performance report.³¹⁸
4. Set up referral tracking for utm_source=chatgpt.com and the other assistant domains in your analytics.⁶
5. Write a prompt set. 30 to 60 questions your ICP asks across the buying stages, in the words sales hears on calls.
6. Take a baseline. Run each prompt several times per engine. Record mentions, citations and which third-party domains are cited.¹⁷

Days 31 to 60: fix your own pages and map the third-party gap

1. List the domains that keep getting cited for your prompts but do not mention you. This is your earned media target list.
2. Write pages that answer evaluation questions directly: pricing model, integrations, security, implementation time, who you are not for. Use first-hand detail.³
3. Fix factual drift. If answers repeat outdated facts about you, find the cited sources and correct or request corrections.
4. Keep structured data accurate to visible content. Do not invent new markup for AI.¹⁴

Days 61 to 90: earn mentions and decide on tooling

1. Pitch inclusion to the comparison and review pages on your list where you are a genuine fit, with accurate facts.
2. Publish one piece of original data that others can cite.
3. Activate customer reviews on the review platforms that appear in your baseline citations.
4. Re-run the baseline with the same prompts and repeat counts. Compare appearance share, not positions.
5. Decide on a tracker. If manual sampling became a burden and the numbers matter, buy one covering your buyers’ engines. If not, keep sampling monthly and fund PR instead.

9. Mistakes to avoid

• Blocking search crawlers while meaning to block training. The most damaging error and the easiest to fix.⁵⁸
• Ignoring Bing. It feeds Copilot.⁹¹¹
• Buying a tracker before fixing access. You pay to measure a problem visible in robots.txt.
• Reporting AI “rank.” Position varies too much between runs to mean anything.¹⁷
• Mass-producing prompt-shaped pages. Google names scaled content abuse as a risk here.³
• Faking mentions. Planted forum posts and self-written reviews are a reputational risk and, per Google, less useful than they look.³
• Treating llms.txt or new schema as the plan. Google says neither is needed for its AI features.²³
• Expecting clicks to replace lost organic traffic. Click rates are lower when an AI summary appears.¹³ Judge the program partly on branded demand and pipeline quality.

10. The verdict

AI search visibility is real and measurable, and most of what drives it is work a competent B2B marketing team already knows. The entry requirements are basic, the best public evidence points to third-party reputation, and measurement needs repeated sampling. The new work is narrow: crawler access, Bing, a prompt set and patience.

Frequently asked questions

Is GEO different from SEO?

Mostly in emphasis, not in fundamentals. Google says its AI features rely on its core ranking and quality systems and have no extra technical requirements.²³ The differences are in measurement (appearance share instead of rank), the higher weight of third-party mentions, and the need to manage AI-specific crawlers.

Should we block GPTBot and other AI training crawlers?

That is a business and legal decision about whether your content may train models. It is separate from search visibility at OpenAI, Anthropic and Google, so you can block training and remain citable.⁵⁸¹⁰ Just make sure the search crawlers stay allowed.

Does adding llms.txt help us get cited?

There is no documented evidence that it does for AI search. Google says Search ignores it,³ and none of the other platforms’ crawler documentation we reviewed mentions it. It may help agents navigate developer docs, which is a narrower use.

How do we know if ChatGPT is sending us traffic?

ChatGPT adds utm_source=chatgpt.com to referral links, so filter your analytics on that parameter.⁶ Expect undercounting, because many buyers read an answer and then visit you later through branded search or direct.

How many prompts should we track?

There is no official number. The SparkToro and Gumshoe research suggests that repeating each prompt many times matters more than a long list run once, because single answers vary so much.¹⁷ Start with 30 to 60 prompts tied to real buyer questions and run each several times per engine.

How long until we see results?

Crawler fixes can register within about a day on OpenAI and Perplexity, by their own documentation.⁵⁷ Google says preview control changes can take from several days to several months to take effect.² Changes driven by earned media take as long as the coverage takes to land, so judge the program over a quarter, not a week.

Sources and notes

1. Source: Aggarwal et al., “GEO: Generative Engine Optimization,” arXiv 2311.09735, accessed Oct 2026. Abstract reports visibility gains of up to 40% on its benchmark and notes efficacy varies by domain. ↩ ↩²

2. Source: Google Search Central, “AI features and your website”, accessed Oct 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

3. Source: Google Search Central, “Optimizing for generative AI features on Google Search”, accessed Oct 2026. Announced on the Search Central blog in May 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶

4. Source: TechCrunch, “Google’s AI Overviews have 2B monthly users, AI Mode 100M in the US and India”, accessed Oct 2026. Figure stated by Sundar Pichai on the Q2 2025 earnings call. ↩

5. Source: OpenAI, “Overview of OpenAI crawlers”, accessed Oct 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰

6. Source: OpenAI Help Center, “Publishers and Developers FAQ”, accessed Oct 2026 via search result text; the page blocked direct retrieval. ↩ ↩² ↩³ ↩⁴

7. Source: Perplexity, “Perplexity crawlers”, accessed Oct 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

8. Source: Anthropic Help Center, “Does Anthropic crawl data from the web, and how can site owners block the crawler?”, accessed Oct 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

9. Source: Microsoft Bing Blogs, “Evolving role of the index: From ranking pages to supporting answers”, accessed Oct 2026. ↩ ↩² ↩³

10. Source: Google Crawling Infrastructure, “Google’s common crawlers”, accessed Oct 2026. Google-Extended is a robots.txt token with no separate user agent. ↩ ↩²

11. Source: Bing Webmaster Blog, “Introducing AI Performance in Bing Webmaster Tools Public Preview”, accessed Oct 2026. Dated February 10, 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

12. Source: Ahrefs, “An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied)”, accessed Oct 2026. Spearman correlations; published May 2025 by a vendor of AI visibility tools. ↩ ↩² ↩³ ↩⁴

13. Source: Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results”, accessed Oct 2026. Browsing data from 900 US adults, March 2025. ↩ ↩² ↩³ ↩⁴

14. Source: Google Search Central, “Introduction to structured data markup in Google Search”, accessed Oct 2026. ↩ ↩²

15. Source: llmstxt.org, “The /llms.txt file”, accessed Oct 2026. ↩

16. Source: Search Engine Roundtable, “Google: No AI System Currently Uses LLMs.txt”, accessed Oct 2026 via search result text. Reports a June 2025 Bluesky post by John Mueller. ↩

17. Source: SparkToro, “NEW Research: AIs are highly inconsistent when recommending brands or products”, accessed Oct 2026. Published January 28, 2026; not peer reviewed; co-author works at an AI tracking vendor. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

18. Source: Google Search Central Blog, “Introducing Search Generative AI performance reports in Search Console”, accessed Oct 2026. Posted June 3, 2026. ↩ ↩² ↩³

19. Source: Microsoft Bing Blogs, “New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare”, accessed Oct 2026 via search result text. ↩

20. Source: Ahrefs, “Pricing”, accessed Oct 2026. List prices; Brand Radar is an add-on. ↩

21. Source: Semrush Knowledge Base, “AI Visibility Toolkit”, accessed Oct 2026. List price; no free trial per the page. ↩ ↩²

22. Source: Semrush, “Plans and pricing”, accessed Oct 2026. Monthly billing list prices. ↩

23. Source: Peec AI, “Pricing”, accessed Oct 2026. The page names ChatGPT, Perplexity and Gemini as tracked engines. ↩

24. Source: Profound, “Pricing”, accessed Oct 2026. ↩ ↩²