Should I trust an agency that guarantees AI citations?

A guarantee assumes someone controls the outcome. In AI search, nobody outside the engine does. Every AI answer passes through four stages:

• Parametric: what the model learned in training, frozen at a cutoff. Nobody changes this quarter's model.
• Access: whether a crawler can fetch and store your page. You and your agency control this.
• Retrieval: whether your page enters the model's context for a given query. The engine decides, and it often rewrites the user's prompt into its own queries.
• Generation: which retrieved pages the model uses and cites. The model decides.

Retrieval is also discretionary. As of October 2026, Google AI Overviews often don't trigger, and ChatGPT decides per query whether to search at all. Google's own documentation says pages must be indexed and snippet-eligible to appear in AI Overviews or AI Mode, and adds that indexing and serving isn't guaranteed. If Google won't guarantee placement on its own surface, an agency can't.

So a guarantee usually means one of three things. The agency misunderstands how engines work. It defines success with a metric it controls. Or it relies on tactics that manipulate engines, which carry real risk. Our guide to GEO claims that don't hold up covers the most common versions.

The work with documented evidence is mostly about removing blocks. An agency can promise to check and fix these:

• Rendering: many AI crawlers don't execute JavaScript, so content that appears only after scripts run may be invisible to them.
• Per-bot robots.txt rules: OpenAI's crawler docs say each setting is independent. GPTBot governs training. OAI-SearchBot governs ChatGPT search.
• Network reachability: a firewall or CDN can block a crawler that robots.txt allows.
• Suppression controls: noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers, sometimes by accident.
• Fact consistency: stating your price or key facts clearly, and the same way everywhere, gives engines a sentence they can quote.

These are promises about work, not results. Fixing access makes you eligible. It does not make an engine choose you. A sound AI SEO service is scoped around deliverables like these.

Most citation promises rest on content tactics: rewriting for AI, adding schema, inserting statistics. The evidence is thin.

A July 2026 critical survey of 45 GEO studies (arXiv 2607.14035) found that no reviewed technique shows a stable, longitudinal, cross-platform causal effect on discoverability. Effects appeared only on content already retrieved.

C-SEO Bench, presented at NeurIPS 2025, found most conversational SEO methods largely ineffective and often harmful to ranking. Ahrefs tracked 1,885 pages that added schema against matched controls and found AI Overviews citations fell 4.6%.

The widely quoted "up to 40% more visibility" comes from a lab engine where sources were already loaded into context. It says nothing about getting found in live search.

Some tactics do move answers: hidden text and prompt injection aimed at AI crawlers. They are illegitimate, and they put your brand at risk with the platforms. If an agency's results look too good, audit your site for risky GEO tactics.

Replace "can you guarantee citations?" with questions that reveal how the agency works.

1. What exactly do you check? Expect specifics: raw HTML versus rendered page, robots.txt per bot, crawler requests through your CDN, snippet directives.
2. Which numbers are measured and which are modelled? Any single AI visibility score averages answers given with and without live retrieval. Ask how theirs is built.
3. What evidence backs each recommendation? Ask for the primary source, not a blog post.
4. Do you report each engine separately? ChatGPT, Perplexity and Google's AI surfaces use different indexes and behave differently.
5. What do you report if nothing changes? A good answer names the work done and what remains outside anyone's control.
6. Do you use hidden text, cloaking or mass-generated pages? The only acceptable answer is no.

Watch for these phrases too: "E-E-A-T score", "domain authority for AI", "publish an llms.txt", and "your rank in Google AI". Google exposes impressions only for its AI features, so no position data exists to check a rank against. Our questions to ask an AI SEO agency go deeper.

Not every guarantee is a problem. An agency can reasonably guarantee deliverables: an audit by a date, every crawler block found and documented, fixes shipped within an agreed window. Those sit inside its control.

Case studies deserve care. Citation counts track brand size closely, and one site's gain rarely separates the tactic from everything else that changed. OtterlyAI's 2026 experiment shows how misleading raw counts can be. Google deindexed both of its mass-generated AI sites, yet Microsoft Copilot kept citing them because Bing kept them indexed. Rising citations on one engine can hide losses on another. Read how to judge GEO case studies before you weigh one.

Start with the parts you can verify yourself. Check whether your robots.txt treats OAI-SearchBot, PerplexityBot and Claude-SearchBot the way you intend, and whether your key facts appear in the HTML before JavaScript runs. The free checker on this site gives a snapshot of whether AI engines name you for up to three buyer questions.

Then compare agencies on scope, method and honesty about limits. Choose the one that tells you what it can't control.