Answer Engines vs Search Engines: What You Control

A classic search engine does three jobs: crawl, index, rank. You type a query, it returns ten links, and you pick one. The ranking is the product.

An AI search engine keeps the index and adds work on both sides of it. Before retrieval, it rewrites your prompt into one or more searches of its own. After retrieval, a language model reads the pages it pulled in and writes a single answer. Citations, if any, are a by-product of that writing step.

That changes what "visible" means. In search, visible means ranked. In an answer engine, visible means your page was fetched, then retrieved for a query you never saw, then used by a model that decides what to say. For the wider channel question, see whether AI search and SEO are one system.

Every answer an AI engine gives passes through up to four stages. Separating them shows where your effort lands.

Two rows deserve attention. The parametric row has no equivalent in search. When an answer engine skips retrieval, the answer comes from training data. If your company launched after the model's cutoff, that answer can't include you (more on that in whether ChatGPT can know a new company).

The generation row is also new. A ranked list shows your title and snippet as written. A generated answer paraphrases, merges and sometimes gets your facts wrong.

Query rewriting

OpenAI's help page says ChatGPT search "rewrites your query into one or more targeted queries" before sending them to search providers. Google describes a query fan-out technique that issues multiple related searches across subtopics. The Gemini API generates its own search queries too.

What it means: the prompt a buyer types is not the query that retrieves your page. Any tool that tests prompts is measuring a proxy for an input nobody outside the engine can see.

Discretionary retrieval

As of October 2026, retrieval fires per query and at the engine's discretion. ChatGPT decides when to search. Claude.ai search is a user toggle. Google AI Overviews often don't trigger at all. So the same question can get a grounded answer one day and a training-data answer the next.

Generation and citation

OpenAI's API returns the URLs a model consulted separately from the URLs it cited, and its docs say the consulted list is often longer. Being retrieved and being cited are different events. A citation count undercounts which pages shaped the answer.

The levers that hold up under scrutiny sit at the access stage. They work by removing blocks, not by adding signals.

• Render the facts that matter in the HTML. Vercel's crawler study found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so client-side content can be invisible to them. Googlebot does render.
• Set robots.txt per bot. OpenAI's crawler docs say each setting is independent. GPTBot governs training. OAI-SearchBot governs ChatGPT search.
• Check the firewall and CDN. OpenAI says eligibility needs robots.txt permission and a host or CDN that allows its search crawler's IP addresses. Since 15 September 2026, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains and free-tier customers, while Search stays allowed.
• Stay index and snippet eligible. Google says a page must be indexed and eligible to show with a snippet to appear as a supporting link in AI Overviews or AI Mode, with no additional technical requirements. Stray noindex, nosnippet, max-snippet:0 or data-nosnippet tags keep content out.

A common failure is a blanket AI block that removes search eligibility along with training. This keeps search open and opts out of training:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

robots.txt can say allow while a firewall returns 403. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away. The ChatGPT specifics are in getting cited by ChatGPT.

Once a page is reachable, the engine takes over. Which pages enter the model's context, in what order, and which ones get quoted are decisions the engine makes.

The research on influencing those decisions is thin and mostly negative. C-SEO Bench (NeurIPS 2025) found most conversational SEO methods "largely ineffective" and frequently harmful to a document's ranking. It also found that moving a document higher in the model's context mattered far more, which no publisher can do directly. A July 2026 critical survey of 45 GEO studies concluded that no reviewed technique shows a stable, cross-platform causal effect on discoverability. Nearly all published GEO effects were measured on pages already inside the model's context.

The same goes for markup. Google says structured data isn't required for generative AI search. Ahrefs' matched-control test found AI Overview citations fell 4.6% after pages added schema. Existing markup does no harm, but it isn't a lever for AI answers.

The parametric stage is fully out of reach in the short term. A new model release can change what an engine says about you overnight, as covered in what a new model release does to brand answers.

Classic SEO still feeds answer engines, but it no longer covers the ground. Ahrefs found that 37.9% of URLs cited in AI Overviews also appeared in the first 10 results (March 2026). Ranking helps, and most citations still come from elsewhere.

Three things change in practice:

1. Access becomes per engine. One crawler (Googlebot) used to be the gate. Now OAI-SearchBot, PerplexityBot, Claude-SearchBot and others each need a path in, and each can be blocked separately.
2. Facts matter more than pages. A model quotes sentences. Stating a specific price, number or one-line definition on a fetchable page gives it a sentence to quote. Contradictions across your own pages give it room to be wrong.
3. Measurement splits by engine. Google's AI report shows impressions only, with no clicks or position. There is no AI rank to track. Check each engine on its own and note whether the answer was grounded.

Use classic search thinking when the question is whether Google can index and rank your page. That work also gates Google's AI Overviews and AI Mode, since both draw on the same index.

Use answer-engine thinking when the question is what ChatGPT, Perplexity, Claude or Gemini say about you:

• Nothing appears at all: check access first. Rendering, per-bot robots.txt, firewall and snippet tags.
• Engines cite other sites instead of yours: being named on those sites is associated with appearing in the answer. Read whether press releases help before spending on distribution.
• Engines get your facts wrong: state the correct facts plainly on fetchable pages and remove contradictions.
• You want a baseline: the free checker on this site shows whether engines name you for three buyer questions. Treat it as a snapshot, not a trend.

Is SEO still worth doing if answer engines replace some searches? Yes. Google's AI Overviews and AI Mode require a page to be indexed and snippet eligible, so classic indexing work is a precondition there. Ahrefs found 37.9% of AI Overview citations came from pages in the top 10 (March 2026). Ranking helps but doesn't guarantee inclusion. Treat SEO as one input, and add per-engine access checks for ChatGPT, Perplexity and Claude.

Does blocking GPTBot remove my site from ChatGPT search? No. OpenAI's crawler documentation says each setting is independent. GPTBot controls whether content may be used for training. OAI-SearchBot controls whether pages can appear in ChatGPT search answers. You can block GPTBot and allow OAI-SearchBot. The reverse is also true: allowing GPTBot does nothing for search visibility. Check your firewall too, since it can block a crawler that robots.txt allows.

Will schema markup or an llms.txt file get me into AI answers? The evidence says no. Google states structured data isn't required for generative AI search, and Ahrefs' matched test found AI Overview citations fell 4.6% after pages added schema. Google says Search ignores llms.txt, and no other engine documents reading it. Existing markup does no harm. It just isn't a lever for AI answers. Spend the effort on rendering and crawler access instead.

Why do different answer engines say different things about my brand? Each engine uses its own crawlers, index, query rewriting and model, and each decides per query whether to search at all. One answer may come from live retrieval and another from training data with an older cutoff. Google says even AI Overviews and AI Mode may use different models and return different links. Check each engine separately instead of relying on one blended number.