fourcents/seo
Visibility

AI Search Optimization

People now ask assistants what they used to type into a search box. Our AI search optimization service makes sure AI systems can reach your content, understand who you are, and have a reason to cite you, and it measures whether they do.

  • AI crawlers allowed and able to read your content
  • Content structured so it can be quoted
  • Citations and referrals from AI tools tracked every week

How AI search differs

An assistant does not return ten links. It composes an answer and may cite a few sources. To be one of them, your content has to be reachable by the assistant's crawler, clear enough to extract a direct answer from, and attributable to an entity it recognises.

Retrieval still starts from search indexes. Google's AI Overviews and AI Mode draw on Google's index; ChatGPT search leans on Bing's index and its own crawler; Perplexity runs its own. That is why AI search optimization is not a replacement for SEO. It is SEO with a different scoreboard: citations and AI referrals rather than positions alone.

What we check first

  • Crawler access. Whether OAI-SearchBot, ChatGPT-User, GPTBot, PerplexityBot, ClaudeBot and Google-Extended are allowed in robots.txt and not blocked by a CDN or firewall rule. Many CDNs block them by default.
  • Rendering. Most AI crawlers do not run JavaScript. Content that only appears after scripts load is invisible to them.
  • Bing indexing. ChatGPT search results overlap heavily with Bing's top results, so a site missing from Bing is missing from ChatGPT. We set up Bing Webmaster Tools and IndexNow.
  • Answer-first structure. Whether each page states its answer near the top in plain sentences an assistant could quote.
  • Entity clarity. Organization and author markup, consistent names, descriptions and profiles across the web.
  • llms.txt. A plain-text guide to your most useful pages. We publish it because it is cheap and harmless, and we are clear that Google has said it has no effect on Search rankings or AI Overviews.

The programme, step by step

  1. Baseline. A panel of 40 prompts per topic is run across Google AI Overviews and AI Mode, ChatGPT search, Perplexity and Copilot. We record who is cited, how your brand is described, and which of your pages appear. This is repeated weekly for the life of the engagement.
  2. Access. robots.txt and firewall rules fixed, Bing Webmaster Tools and IndexNow set up, sitemap and lastmod dates corrected, llms.txt published.
  3. Entity work. Organization or LocalBusiness schema with sameAs links, consistent descriptions on every profile you control, named authors with credentials on content that makes claims.
  4. Answer-first rewrites. The pages that should be cited get openings that answer, headings that ask the question, definitions and tables an assistant can lift, and dates that show the page is current.
  5. Original material. Data, tools or findings nobody else has, because generic overviews are not cited; the assistant can write those itself.
  6. Mentions. Ahrefs' 2026 study of 75,000 brands found unlinked brand mentions correlate with AI citations about three times as strongly as backlinks. Digital PR on Growth and Scale is aimed at mentions, not only links.
  7. Measurement. Citation share on the panel, AI referral sessions and leads in GA4, and crawler activity in server logs, in the monthly report.

What we need from you

  • Access to robots.txt and your CDN or firewall settings, or a developer who has it.
  • Search Console, Bing Webmaster Tools (we set it up if you do not have it) and GA4 access.
  • A named author for content, with a short biography, and agreement on how the company should be described everywhere.
  • A subject expert for the facts, figures and opinions that make a page worth citing.
  • A decision on training crawlers: allowing retrieval agents is our standing advice; whether to allow GPTBot and Google-Extended for training is your policy call, and we implement either.

What it can and cannot achieve

It can make the site retrievable by every major assistant, raise the share of prompts where you are cited on a defined panel, and produce measurable referrals and leads from ChatGPT, Perplexity and Copilot. On our own network, the first inbound lead on a new site arrived from ChatGPT two days after launch, before Google had shown a single impression.

It cannot guarantee a citation in a specific answer. AI answers are assembled per query, per user and per session, so visibility is measured by repeated sampling, not a fixed rank. It cannot give you click data for AI Overviews either: Search Console's AI performance report, launched in June 2026, shows impressions for AI features but not clicks. And it cannot reverse the fall in clicks on informational queries; the realistic goal there is to be the cited source and to win the commercial queries that still send visits.

How it is measured

  • Citation share: the share of panel prompts, per assistant, where your site is cited, tracked weekly against the baseline.
  • How you are described: whether the assistant's description of the company is accurate and matches the entity work.
  • AI referrals: sessions from chatgpt.com, perplexity.ai, copilot.microsoft.com and utm_source=chatgpt.com in GA4, and the leads they produce.
  • Crawler activity: fetches by OAI-SearchBot, ChatGPT-User, PerplexityBot and ClaudeBot in server logs, which show which pages are actually being retrieved.
  • AI feature impressions in Search Console's generative AI report.

Common failure modes

  • Blocked without knowing. A bot-protection setting on the CDN challenges every AI crawler. The site is invisible to assistants and nobody chose that.
  • Treating llms.txt as the strategy. It is a courtesy file. Google does not use it, and no major provider has committed to it as a citation signal.
  • Optimising for a single test run. One prompt, one answer, one screenshot proves nothing. Only a repeated panel over weeks shows movement.
  • Content written for assistants rather than people. Assistants cite the pages people trust. Stilted, keyword-heavy text loses on both counts.
  • Inconsistent entity data. Three different company descriptions, two addresses and an old logo make the assistant unsure who you are.

What we cannot promise

Nobody can guarantee a citation in an AI answer. The systems are opaque and change often. We improve the odds with things that are known to matter, we measure on a fixed panel so you can see whether it is working, and we report the referrals and leads with dates. The longer version of this page, with the research behind it, is on the AI search optimization overview.

Deliverables

DeliverableWhat it isCadence
AI-readiness auditCrawler access, rendering, Bing indexing, answer structure, entities and llms.txtAt start
Prompt panel baseline40 prompts per topic across AI Overviews, ChatGPT, Perplexity and Copilot, with who is citedAt start, then weekly
Access and indexing fixesrobots.txt, CDN rules, Bing Webmaster Tools, IndexNow, sitemap, llms.txtMonth one
Answer-first rewritesPriority pages restructured so answers are extractable and datedMonthly
Entity and schema workOrganization, author and content markup, consistent profilesAt start, then as needed
AI visibility reportCitation share, referrals, leads and crawler activityMonthly

How it runs

  1. Audit accessConfirm AI crawlers can reach and read the site, and that Bing has it indexed.
  2. BaselineRun the prompt panel and record who is cited today.
  3. Restructure and clarifyMake answers explicit and quotable; tell machines who you are.
  4. TrackRe-run the panel weekly; report citations, referrals and leads monthly.

Which plan includes it

Starter $2,200Growth $4,500Scale $8,500
AI-readiness audit
Crawler access and llms.txt
AI-referral tracking in GA4
Everything in Starter
Weekly prompt panel
Monthly answer-first rewrites
Entity and brand-mention work
Everything in Growth
Original data projects
Ongoing mention monitoring

Compare all plans →

Questions

Is AI search replacing Google?

Not yet, but a growing share of research starts in an assistant, and Google itself now answers many queries with AI Overviews. The work overlaps heavily with good SEO, so it is rarely wasted.

Should we block AI crawlers?

Blocking retrieval agents (OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot) removes you from the answers your customers read. Blocking training bots (GPTBot, Google-Extended) is a policy choice. We help you decide per crawler and implement it.

What is llms.txt?

A plain-text file listing your most useful pages for language models. It is a proposed convention, cheap to add and harmless. Google has said it does not use it, so we treat it as a courtesy, not a tactic.

Can you guarantee we appear in ChatGPT?

No, and nobody honestly can. We improve the factors that are known to matter and measure the result on a repeated panel.

How do you track traffic from AI tools?

GA4 segments on referrer (chatgpt.com, perplexity.ai, copilot.microsoft.com) and on utm_source=chatgpt.com, which ChatGPT adds to many links, plus server logs for crawler fetches.

How long before we see citations?

Access fixes can show in crawler logs within days. Movement on the prompt panel usually takes one to three months, and depends on how much original, specific material the site has to cite.

Related

Find out what is holding your site back.

A 30-minute call with the person who will do the work, after we have crawled your site. No slides, no pressure.

Free auditBook a call