AI Search Optimization

Getting found and cited by AI search

AI search optimization is the work of making a website readable and quotable by the assistants people now ask instead of typing into a search box — ChatGPT, Google’s AI Overviews, Microsoft Copilot, Gemini and Perplexity. It overlaps with SEO but is not the same job: these systems lift short passages out of pages, so the work is server-rendered HTML that a crawler can read without running JavaScript, structured data that states plainly what the business is, content written answer-first so a single passage stands on its own, and robots rules that actually let the right crawlers in. Nobody can buy a citation, but a business can be the clearest and most machine-readable source on the questions its customers ask.

Start a projectAll services
AI Search OptimizationService 08 / 08

What is included

  • Crawler access report
  • Robots.txt rewrite
  • Server-rendered HTML
  • Validated structured data
Who it is for 4 audiences

Who this is for

  • 01

    Businesses people ask about by category

    Restaurants, venues, contractors and clinics. When someone asks an assistant where to eat in Kalamazoo or who books event space, a name gets returned. Right now it is probably not yours.

  • 02

    Sites built as JavaScript applications

    React and single-page builds where the content only exists after scripts run. Google renders those. Most AI search crawlers do not, so they see a near-empty shell of a page.

  • 03

    Companies already investing in SEO

    You rank fine, traffic is flat or falling, and the clicks are going to an answer box instead. The pages need restructuring for retrieval, not another round of keyword work.

  • 04

    Anyone being described incorrectly

    Assistants confidently repeat an old address, a service you dropped, or a competitor’s detail attached to your name. Correcting that is entity work, and it is fixable.

Problems 6 signs

Problems this solves

  • Your content only appears after JavaScript executes, so crawlers that do not render pages see headers, footers and almost nothing else.
  • A CDN or firewall started blocking AI crawlers by default during a security update, and nobody made a deliberate decision about it.
  • Asking ChatGPT or Perplexity a question your business exists to answer returns three competitors and no mention of you.
  • Assistants describe you inaccurately — outdated hours, a former location, or a service line you stopped offering two years ago.
  • Every page opens with a paragraph of throat-clearing before the actual answer, leaving no clean passage worth quoting.
  • Structured data was added by a plugin, so it is generic, duplicated across pages, or throwing validation errors nobody has looked at.
Deliverables 8 items

What you get

Scope
Set in a written proposal, delivered within 48 hours.
Who builds it
Sahil Sandhu, start to finish — one developer, no hand-offs.
  1. 01A crawler access report covering robots.txt, CDN and firewall rules, listing each relevant bot — OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot, Bingbot, Googlebot — with a recorded decision for each rather than an accidental default
  2. 02A rewritten robots.txt that separates crawlers used for model training from the search crawlers that produce citations, so you can decline one without losing the other
  3. 03Server-rendered HTML for every page that matters, verified by fetching the site with JavaScript disabled and reading what actually comes back
  4. 04Structured data written by hand and validated: Organization or LocalBusiness, Service, Product, Article and FAQPage where real content supports it, with sameAs links tying your profiles to one entity
  5. 05Answer-first rewrites of priority pages — a direct definition in the opening lines, headings phrased as the questions people ask, and passages that make sense quoted alone
  6. 06An entity consistency pass so your name, description, category, address and hours match across the site, Google Business Profile, directories and social profiles
  7. 07An llms.txt file and clean plain-text versions of key pages, published as a low-cost convention rather than sold as a ranking factor
  8. 08A baseline prompt test: the real questions your buyers ask, run against ChatGPT, Gemini, Copilot and Perplexity, with the answers and cited sources recorded so later runs can be compared
Process 4 steps

How the work runs

  1. 01

    See the site as a machine does

    We fetch your pages without JavaScript, check robots.txt and firewall rules bot by bot, validate existing schema, and ask the major assistants about your business to record exactly what they currently say and cite.

  2. 02

    Open access and fix the plumbing

    Blocked crawlers are unblocked deliberately, pages that depended on client-side rendering are moved to server-rendered HTML, and hand-written structured data replaces whatever generic markup was there before.

  3. 03

    Rewrite for retrieval

    Priority pages get restructured so each section answers one question completely and early. Retrieval happens at passage level, so a paragraph that only makes sense after reading the two above it never gets used.

  4. 04

    Re-test and adjust

    The same prompts get run again on a schedule. We track which pages get cited, where a competitor is being quoted instead, and which questions have no good source yet — then write the page that answers them.

Stack & timeline 6 tools

Technology and timeline

Technology

  • Next.js
  • React
  • TypeScript
  • Node.js
  • Vercel
  • Shopify/Liquid

Timeline

The audit and prompt baseline typically take one to two weeks, and the access and structured data fixes usually ship in the few weeks after that, since they are engineering changes rather than content work. What happens next is outside anyone’s control: assistants have to recrawl and reindex before changes register, which often takes days to weeks for search-connected answers and longer for anything drawing on older data. Most engagements run in three- to six-month blocks so the same prompts can be re-tested more than once.

Pricing 5 packages

What it costs

This is quoted in two parts. The audit and technical implementation are a fixed project price once the site has been reviewed, because the scope is knowable — a set number of pages to server-render, one robots and firewall configuration, a defined schema build. Ongoing prompt monitoring and content work is an optional monthly retainer sized by how many questions you want tracked and how much writing is included, and it is never tied to promised placement in an AI answer, because no such promise can be honestly made.

Starting prices cover the scope each package describes. After a free call, your written proposal fixes the price within 48 hours.

Monthly plans are a flat price and run month to month: cancel before any billing date.

See all packages
Related work 3 projects

Related work

FAQ 6 questions

Frequently asked questions

Last reviewed

01

Is this just SEO with a new name?

They share a foundation and diverge from there. Both need a crawlable, fast, well-structured site. But search engines rank whole pages while assistants retrieve passages, so the writing target changes: a self-contained answer in the first two sentences matters more than keyword placement, and structured data does more work because it removes guesswork about what your business is.

02

Can you guarantee ChatGPT will recommend my business?

No. There is no ranking to buy, no submission form, and the same question can produce different answers to different people on different days. What can be committed to is the work: making your site machine-readable, unblocking the crawlers that generate citations, stating your business clearly in structured data, and answering the questions buyers actually ask better than anyone else does.

03

Should I block AI crawlers to protect my content?

That depends on which crawler, and the distinction matters. Some bots gather text for model training and give nothing back. Others fetch pages live to answer a question right now and cite the source. Blocking the second group removes you from AI answers entirely. We separate them in robots.txt so the choice is yours and made on purpose.

04

Does llms.txt actually do anything yet?

Honestly, not much on its own. It is a proposed convention for a plain-text file summarising your site for language models, and no major assistant has publicly committed to reading it. It costs almost nothing to publish and may matter later, so we include it. It is not the load-bearing part of this work, and anyone selling it as one is overselling.

05

Do Google’s AI Overviews need something different from normal SEO?

Mostly the same foundation, with a different emphasis. AI Overviews draw on Google’s regular index through Googlebot, so being indexed well still comes first. The difference is what gets pulled into the summary: short, direct, factually clean passages with clear headings. Pages that answer a question in one paragraph get quoted far more than pages that circle it.

06

How do I know whether any of this worked?

Through repeat prompt testing rather than a ranking chart. We record what each assistant says about your category and who it cites before the work starts, then run the identical prompts again on a schedule. You see the answers change or you do not. Server logs also show whether the AI crawlers are now reaching your pages at all.

Related services

Ready to talk about AI search optimization?

You will hear back within one business day, with a written proposal — timeline, approach and a fixed price — inside 48 hours.

Start a projectCall (269) 929-6895