GEO Strategies: 20 Popular Tactics Checked Against the Evidence

The GEO strategies that hold up against primary sources are about access and accuracy: let AI crawlers reach your pages, stop blocking your own snippets, and make your published facts agree. Most of the other popular strategies, from schema and llms.txt to "write for AI" formatting and automated distribution, are unmeasured, measured as having no effect, or contradicted in writing by the platform they claim to influence.

We checked 20 commonly sold tactics against the vendors' own documentation and the controlled studies. Here is the scoreboard, how we graded it, the detail behind each verdict, and a 30-day plan built only from what survived.

The scoreboard

GEO strategy	Verdict	What decides it
Allow AI search crawlers in robots.txt, bot by bot	Holds up	OpenAI and Meta state the consequence in writing
Make sure your firewall or CDN lets them through	Holds up	OpenAI makes it half of eligibility
Remove accidental snippet blocks	Holds up	Google and Bing apply them to AI answers
Server-render the content that matters	Holds up	ChatGPT's and Claude's crawlers didn't run JavaScript (Vercel, December 2024)
Make your facts agree everywhere	Sensible, untested	Fully in your control; no controlled test
Be the most relevant answer to the question	Holds up	The most reproducible factor in the research
Add schema markup for AI	No measured effect	1,885-page matched test found no lift
Publish an llms.txt file	No documented reader	Google says Search ignores it
Add FAQ schema or FAQ blocks for AI	Retired / no measured effect	FAQ rich results ended May 7, 2026
Raise your E-E-A-T or domain authority	No such score	Google: E-E-A-T isn't a ranking factor
Chunk content into short blocks	No evidence	Google says unnecessary; Microsoft recommends structure; no test
Rewrite in an "AI style"	No effect in lab tests	C-SEO Bench: ineffective or harmful
Refresh dates to look fresh	Unresolved	Self-declared dates prove nothing
"GEO boosts visibility up to 40%"	Lab result	Measured on a benchmark, not live search
Allow GPTBot to appear in ChatGPT	Wrong bot	GPTBot is for training; OAI-SearchBot is for search
Track your rank inside AI answers	No ground truth	No platform exposes a position
Video transcripts as an LLM source	Unmeasured	No engine documents it, no test either way
Automated press-release syndication	Link spam when done for links	Google names optimized-anchor links in distributed press releases
Mass-produced AI pages	Spam risk	Google's scaled content abuse policy
Bot traffic to force rankings	Policy violation	Google's machine-generated traffic policy
Seeding mentions on forums and Reddit	Discounted	Google: inauthentic mentions aren't "as helpful as it might seem"

How we graded each GEO strategy

An AI answer that cites the web passes three stages:

1. Access. A crawler must be allowed to fetch your page, and your firewall must let it through.
2. Retrieval. For each question, the system runs its own searches and picks candidate pages. OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries."
3. Generation. The model decides which of those candidates to use and cite.

Nearly all experimental GEO evidence sits in the last stage. The July 2026 critical survey of 45 GEO studies sums it up: "already-retrieved content can causally alter its citation or use, but no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability." A tactic that only works once your page is already in the candidate set cannot get you into it.

So every tactic above faced three questions: who documents the mechanism, the platform or a case study; whether the claim is subtractive ("remove what blocks you") or additive ("add this and you will be favored"); and whether you could tell if it failed.

GEO strategies that hold up

1. Allow the AI search crawlers, one by one

Each AI company runs separate bots for search, training and user-triggered fetches, and each is a separate robots.txt rule. OpenAI says sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers" (OpenAI crawler documentation). Meta is the only vendor to state a positive effect: "Allowing Meta-WebIndexer in your robots.txt file helps us cite and link to your content in Meta AI's responses" (Meta's crawler documentation).

A robots.txt that stays in AI search while opting out of OpenAI's training crawler looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: Meta-WebIndexer
Allow: /

User-agent: GPTBot
Disallow: /

That split is allowed by design: OpenAI says "each setting is independent of the others." Robots.txt is necessary, not sufficient, though. Perplexity says its user-triggered fetcher "generally ignores robots.txt rules" (Perplexity crawler docs), and OpenAI says robots.txt rules "may not apply" to ChatGPT-User.

2. Make sure your firewall agrees with your robots.txt

OpenAI's ChatGPT search help page states eligibility as a conjunction: "allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses." Robots.txt says allow, the firewall says 403, and no dashboard tells you.

CDN defaults can change under you. Cloudflare's July 2026 changelog says that from September 15, 2026, new domains get defaults where "bots classified as Training or as Agent are blocked on pages that display ads." Our breakdown of the Cloudflare AI crawler default covers who is affected.

3. Remove accidental snippet blocks

Google's robots meta tag documentation says nosnippet "will also prevent the content from being used as a direct input for AI Overviews and AI Mode," and max-snippet limits it. Bing says data-nosnippet content is "excluded from snippets and AI summaries." These directives often arrive by accident, copied from a template or a plugin.

Google's AI optimization guide adds one more gate: a site "must be included in Search generative AI features in Search Console to be eligible." Check that setting in Search Console for every property you own.

4. Server-render the content that matters

Vercel's December 2024 crawler study found ChatGPT's and Claude's crawlers fetch JavaScript files but "don't execute them. They can't read client-side rendered content." Google's Gemini and AppleBot do render JavaScript.

If your prices, plan names or product descriptions appear only after scripts run, two of the largest AI crawlers never see them. Press Ctrl+U on a key page and search for the fact; if it is missing from the source, it is missing for them.

5. Make your facts agree everywhere

If your pricing table says $49, your FAQ says $59 and an old press release says $39, an AI system can repeat any of the three. No controlled study has measured the effect of fixing that, so we label it untested. It is also fully in your control, it is the problem customers actually complain about, and the fix has no downside.

6. Be the most relevant answer to the question

The survey above found "topical relevance and context position are the most reproducible levers," and a 252,000-trial study found "topical relevance and list position are the biggest drivers of being cited first." You cannot set your position in a model's context. You can write the page that answers the question, and Google says you need not chase "long-tail" variants to do it, because "AI systems can understand synonyms."

GEO strategies that don't hold up

Schema markup for AI

Google says "Structured data isn't required for generative AI search." Ahrefs tracked 1,885 pages that added schema against matched controls: no lift on ChatGPT or AI Mode, and a small, statistically significant 4.6% drop on AI Overviews. Whether schema helps a page with zero citations is untested, not refuted; details in our guide to schema markup for AI.

llms.txt

Google says Search "doesn't use" AI text files and ignores them, and no other AI search engine documents reading them. Ahrefs' server logs found 97% of sites with a valid llms.txt got zero requests in May 2026. It is a reasonable file for developer docs; see what is llms.txt.

FAQ schema and FAQ blocks

Google stopped showing FAQ rich results on May 7, 2026 (Search documentation changelog). FAQ content still helps readers; the markup no longer does anything visible, and the format has no measured AI effect. More in our post on FAQ page SEO.

E-E-A-T scores and domain authority

Google says "E-E-A-T itself isn't a specific ranking factor" (helpful content guide), and its ranking systems guide never uses the term "domain authority," which is a Moz metric. The checkable parts behind E-E-A-T are still worth doing; see E-E-A-T for AI search.

Chunking and "writing for AI"

Microsoft takes the other side on structure. Its Microsoft Advertising blog post of October 8, 2025 says assistants break content down "into smaller, structured pieces" and advises clear headings and modular layouts, with no data behind it.

Google's guide says "you can ignore tactics like 'chunking' content" and "You don't need to write in a specific way" for generative AI search. The 252,000-trial study found "formatting-only edits have little impact." C-SEO Bench went further: "most current C-SEO methods are not only largely ineffective but also frequently have a negative impact."

Refreshing dates

Freshness is unresolved, not refuted: the 252,000-trial study found a recent timestamp "helps consistently" when two sources sit side by side in a model's context. But a date you change without changing the content is self-declared, and bumping it makes your own before-and-after measurements meaningless. Update pages when facts change; our post on content freshness has the detail.

"GEO boosts visibility by up to 40%"

The figure comes from the original GEO paper (KDD 2024), measured on GEO-bench, a benchmark of queries paired with the web sources to answer them. That tests what a model does with sources it has been handed, not whether a page gets found in live search.

"Allow GPTBot to show up in ChatGPT"

GPTBot governs training. Disallowing it "indicates a site's content should not be used in training generative AI foundation models"; ChatGPT search inclusion is OAI-SearchBot's job. Blocking GPTBot does not remove you from ChatGPT search, and allowing it does not add you.

Tracking your rank inside AI answers

No AI platform exposes a position for you to track. Google says "Third-party tools don't have access to our internal ranking data" (Google on third-party SEO tools), and Bing's own AI citation report says its counts reflect "how often pages are cited, not page importance, ranking, or placement" (Bing Webmaster Tools). Measure whether you are mentioned, cited and described correctly.

AI automation tactics, fact-checked

A popular genre of "AI SEO automation" playbook pairs reasonable production advice with visibility claims. Here is how those claims hold up against Google's spam policies (last updated 2026-08-28) and the vendor docs.

• Video transcripts as an LLM source: unmeasured. No engine documents treating transcripts as a distinct source, and no test exists either way. The question you can answer is whether a crawler can reach the text: a transcript loaded by a JavaScript player or hosted on a third-party domain is not text on your page.
• Automated press-release syndication for links: link spam. Google's policy names "links with optimized anchor text in articles, guest posts, or press releases distributed on other sites" and "using automated programs or services to create links to your site." A real release to outlets that cover you is ordinary PR.
• Mass-produced AI pages: scaled content abuse. The policy names "using generative AI tools or other similar tools to generate many pages without adding value for users." AI-assisted content is fine; Google's generative AI content guidance asks you to "manually factcheck and review all AI-generated content" first.
• Bot traffic to force rankings: a policy violation. Google's machine-generated traffic policy covers "sending automated queries to Google," which is what search-and-click bots do. This is not a strategy with a risk profile to weigh.
• Seeding mentions on forums: discounted. Google says "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem." Real participation where your customers talk is fine; manufactured mentions are what this sentence is about.

How to vet a GEO strategy or agency

1. Ask who documents the mechanism. A platform's own documentation beats a case study. A case study without a control group is a correlation with a narrator.
2. Ask whether it is additive or subtractive. "Remove the thing blocking you" is where the documented levers are. "Add X and AI will prefer you" has almost no surviving evidence.
3. Ask how you would know it failed. If the promised metric is a rank inside AI answers, nobody can verify it, and an unverifiable tactic can be sold forever.
4. Ask whether measurements separate search on from search off. An answer from a model's training data will not change when you fix a page; an answer with live search can.

A 30-day GEO plan built on what holds up

1. Week 1, access. Resolve robots.txt for OAI-SearchBot, PerplexityBot, Claude-SearchBot, Meta-WebIndexer and Googlebot. Request your top ten pages as each bot and record any 403, 429 or challenge page.
2. Week 2, suppression and rendering. Search page sources and headers for noindex, nosnippet, max-snippet and data-nosnippet. Check that prices and product facts are in the raw HTML.
3. Week 3, facts. List your ten core facts and make every page, schema block and third-party profile state them identically.
4. Week 4, measure. Ask ChatGPT, Perplexity, Gemini, Claude and Google the questions your customers ask, note which answers used live search, and log every wrong fact as a fix target.

99Visibility's AI visibility audit runs the week 1 access check on your site's main URL and the week 4 measurement for you: it checks robots.txt for 18 AI and search crawlers and firewall access for each one with its own user agent, and it queries ChatGPT and Claude through their APIs with web search on, labelling each answer as searched or from memory, Perplexity with web search on, Gemini with Google Search grounding, and Google AI Overviews and Google AI Mode from Google's live results. New to the field? Start with what is GEO.

Frequently asked questions

How can I optimize my content for GEO?

Make sure AI crawlers can reach it, put the important facts in the raw HTML, and state those facts the same way everywhere. Then write the page that answers the question best. No formatting trick, markup or special file has evidence of making AI cite a page more.

Is SEO dead now with AI?

No. Google says "the best practices for SEO remain relevant for AI features in Google Search," and AI Overviews only draw on pages that are indexed and eligible for a normal snippet. Crawlability, indexing and relevance still decide whether you are in the running.

How do I choose a GEO agency?

Ask for the primary source behind every tactic they propose, and ask how you will know if it failed. Be wary of anyone who promises a rank inside AI answers, sells schema or llms.txt as citation levers, or skips crawler access and firewall checks.

Is AI SEO automation safe?

Automating drafts, repurposing and monitoring is fine if a person fact-checks what gets published. Automating links, mass pages or search traffic is not: Google's spam policies name all three.

Which GEO strategy should I start with?

Access. Check robots.txt for each AI search crawler and request your pages as each one to catch firewall blocks, because a page AI cannot fetch cannot be cited, whatever else you optimize.

To see which AI crawlers can reach your site and what AI says about your brand, run a free 99Visibility audit.