AI search visibility, 30 questions answered.

SEO for AI is the practice of structuring a site so large language models such as ChatGPT, Perplexity, Claude and Google AI Overviews can cite it accurately. It covers JSON-LD schema, entity clarity, review signals, citation-friendly content, and third-party mentions in sources the models index. The goal is to be the business the model names when a user asks, for example, who fits bespoke kitchens in Bristol. Traditional SEO targets a ranked list of blue links. AI search targets a single-sentence answer with a source.

Regular SEO optimises a page to rank on a list. AI search optimises the same page to be quoted inside an answer. The underlying work overlaps, fast site, clear structure, real reviews, but the priorities shift. You need machine-readable facts in JSON-LD, consistent entity data across the web, and content that answers one question per block. Keyword density matters less. Clean data, citations from third parties, and structured answers matter more. A page that ranks third on Google can still be the source an AI quotes. See how this compares to keyword tools.

No. Google AI Overviews pulls from the same index that feeds the blue links, so existing SEO still earns you a seat at the table. Keep your Google Business Profile current, keep earning links, keep publishing useful pages. AI work sits on top. You are adding schema, tightening entity consistency, and writing answer-shaped content that a model can lift without editing. If you stop Google SEO, you lose the base layer the AI models draw from.

UK and US small businesses with a working website that want AI search traffic. Trades, shops, clinics, consultancies, studios, SaaS under 20 staff. It fits best if you already rank on page one or two of Google for your main service terms and want to defend that traffic as search shifts to AI answers. It is less useful if your site has no content, no reviews, and no Google Business Profile, in which case the basics come first.

Schema and on-page fixes are crawled within days. Most sites see Google AI Overviews changes in 2 to 6 weeks. ChatGPT and Perplexity update on different cycles, Perplexity can reflect changes in under a week, ChatGPT often takes 4 to 8 weeks because it mixes live web search with older training data. Review signals and third-party mentions take longer, 30 to 90 days, because they need new content to appear on other sites the models trust.

Both use live web search on top of a base model. Perplexity leans on a smaller set of trusted sources, editorial sites, Reddit, Wikipedia, review platforms, and your own schema. ChatGPT search uses Bing plus its own crawler. Both favour clear entity data, recent content, and third-party mentions over marketing copy. A business named in a regional news article, a directory, and its own schema tends to win over a larger competitor that only has a homepage and a logo.

Not directly. ChatGPT search uses Bing as its live index, then layers in its own crawler and training data. Google AI Overviews uses Google. So the rule is simple, get indexed well in both. Submit a sitemap to Bing Webmaster Tools, not only Google Search Console. Check your site renders in both engines. Many small sites rank fine on Google and invisibly on Bing, which blocks half of AI search on day one.

Run the free checker at /checker.html. Enter your URL, it queries ChatGPT, Perplexity, Claude and Google AI Overviews for 10 prompts a real customer would type, then shows which models mention you and which mention a competitor. You can also ask the models yourself, open each one, type who offers plumbing emergencies near Clifton, Bristol and see who gets named. Screenshot the result. Repeat every 30 days to track movement.

Perplexity refreshes live on every query, so a schema change can show up in hours. Google AI Overviews typically reflects a site change within 1 to 3 weeks, in line with Googlebot re-crawling. ChatGPT search refreshes per query, but the training data layer updates in bulk every few months, which is why some stale facts linger. Claude uses web search and its training cutoff, currently January 2026, so live queries reflect current pages within days.

No. None of the major models sell citation placement in their answers, and buying reviews or fake mentions breaks FTC and ASA rules. What you can pay for is adjacent, legitimate PR in publications the models trust, directory listings, review platform presence, and technical work on your own site. If a service promises paid AI citations in ChatGPT or Perplexity, treat it as a red flag. The models do not have an ad product for this yet.

JSON-LD is a small block of JSON in your page head that tells machines what the page is about in plain terms, this is a local business, the name is X, the opening hours are Y, these are our services. AI models read it before the visible copy because it is unambiguous. A page without schema forces the model to guess from prose, which it often gets wrong. With schema, you control the facts the AI quotes, name, address, phone, services, FAQs, reviews.

For most small businesses, six types cover 90 percent of the work. LocalBusiness or Organization on the homepage, Service on each service page, FAQPage on any page with questions, Article on blog posts, Product on ecommerce items, and Review or AggregateRating where honest reviews exist. Optional but useful, Person for a named founder and BreadcrumbList for navigation. The free schema pack at /free/schema-pack/ ships copy-paste templates for all six.

Inside the head of each page, wrapped in a script tag with type application/ld+json. On WordPress, use the header injection box in your theme or a plugin such as Rank Math or Yoast. On Shopify, edit theme.liquid. On Squarespace and Wix, use the code injection panel in site settings. Each page should have its own schema that matches that page, homepage gets LocalBusiness, a service page gets Service, a blog post gets Article.

Use Google's Rich Results Test and Schema.org validator before you publish. Paste the page URL, it flags missing required fields and syntax errors. For local changes, use the Live URL option so you see the version on staging, not the cached one. If you are editing a theme file, duplicate it first. Schema errors never break a site visually, they just get ignored by the models, so the downside of a mistake is quiet, not catastrophic.

Yes. Google Business Profile feeds Google Maps and part of AI Overviews, but it does not feed ChatGPT, Perplexity or Claude. Those models read your website. Without schema, your site leaves them guessing. The two sources also need to agree, same business name, address, phone, hours. When the models see a mismatch between your GBP and your schema, they either pick one at random or ignore both. Keep them identical.

Yes, and it is often the cheapest customer acquisition channel a sole trader has. A plumber, electrician, joiner or decorator with 30 reviews, a clean schema block, and a page per service area can be the name ChatGPT gives when someone asks for a recommendation. Competitors are slow to act, which is the opening. Start with the workbook at /free/20-questions/, spend an afternoon on schema and services pages, then track citations monthly.

Yes. Review count helps, but it is not a hard gate. We see citations for businesses with 5 to 10 reviews on Google, especially in low-competition local markets. What matters more is that reviews exist on at least two platforms, Google plus one other, such as Trustpilot, Yelp, Checkatrade or an industry-specific site. A single-platform review profile looks thin to the models, two or three platforms looks real.

No, a blog is optional. Service pages with clear FAQs often outperform a blog for citation, because they answer a direct question. If you do blog, write one post per real customer question, not keyword-stuffed filler. A plumbing business with 8 honest blog posts answering things such as how much does a power flush cost in 2026 will get cited more often than one with 80 thin posts. Quality per post matters, volume does not. The SEO for AI blog is our own worked example.

Slightly. A shopfront has a fixed address, so LocalBusiness schema with geo coordinates does most of the work. A service-area business covers a list of towns or postcodes, so you use the areaServed property and create a service-area page per town. Both need consistent Google Business Profile data. Service-area businesses tend to win on intent queries, plumber near me, while shopfronts win on mapped queries, best coffee in Soho.

Yes, with a different schema mix. Product schema on every product page, Review or AggregateRating where genuine reviews exist, Organization on the homepage, and FAQPage on category or buying-guide pages. ChatGPT and Perplexity cite ecommerce sites when a shopper asks for a product recommendation, so clear specs, honest pricing, and review data decide who gets named. Amazon often wins generic queries, but niche Shopify sites can win on specialist queries where Amazon has thin data.

The workbook is a 22-page PDF with 39 specific actions you take on your business website, then it gives you the exact schema, page copy, and FAQ content to paste into your site. It covers LocalBusiness, Service, FAQ, and Review schema, plus a prompt pack for testing your own citations in ChatGPT, Perplexity, Claude and Google AI Overviews. It is designed for a small business owner to complete in one weekend. Get it at /buy/workbook.

The workbook is the $10 PDF, you work through the fixes with our templates over a weekend. The $25 Toolkit is the same methodology as a guided 18-step web flow, the prompts and schema blocks are generated with your business details already inside, with save and resume. If you are confident with copy-paste HTML, the workbook gets you there. If you want to be walked through it step by step, the Toolkit is the easier path. Get it at /buy/toolkit.

Both are supported. Squarespace accepts JSON-LD in the code injection panel under site settings, per-page or site-wide. Wix accepts it in the custom code section. Some older Wix templates strip custom scripts, in which case you use the Wix SEO app to add schema at page level. The workbook and Toolkit include platform notes, so you know exactly where to paste each block. Shopify, WordPress, Webflow, Ghost, and hand-coded HTML are also covered.

Track three things. First, re-run /checker.html every 30 days and log which models cite you. Second, check Google Search Console for impressions on AI Overviews, now broken out as a separate row. Third, watch your referral traffic from chatgpt.com, perplexity.ai, and google.com with the srsltid parameter that flags AI Overviews clicks. Movement on any one of the three within 60 days is a positive signal. Movement on all three by 90 days means the work is compounding.

Three things to check. One, is your schema actually live, run Google's Rich Results Test on the URL. Two, is your site indexed in Bing, not only Google, without Bing you lose ChatGPT. Three, do your Google Business Profile and your schema agree on name, address, phone. If all three are clean and nothing has moved, the gap is usually third-party mentions, your competitors are in directories or editorial sources you are not. That is where the next month of work goes.

Most small businesses benefit from a refresh every 6 to 12 months. The models change how they weight sources, Google rolled out three AI Overviews updates in the last year alone, and your competitors are also acting. A re-audit at month 12 tends to find 10 to 15 new fixes, half technical, half content. If you want help applying any of the new fixes between checks, the $25 Toolkit walks you through the same 18 steps as a guided web flow.

Yes. The workbook is written so a developer can work from it without extra explanation. Each fix includes the exact schema block to paste, where in the code to paste it, and a validation step. Most developers finish the full action list in 4 to 8 billable hours. Forward the PDF, that is the full brief.

Yes. 7-day refund window on the workbook and the Toolkit, no questions asked. Email [email protected] with your order number and we process refunds within 2 working days.