SEO, AEO & GEO
Generative Engine Optimization (GEO): A Practical Playbook
A sourced GEO playbook for small businesses: AI crawler settings, what the GEO study found, answer-first pages, schema, llms.txt and measuring AI visibility.
By Hiba Darwish · Published · 12 min read
Generative engine optimization (GEO) is the work of making your website easy for AI search tools, such as Google AI Overviews, ChatGPT search and Perplexity, to find, understand and cite. For most small businesses, GEO is mostly SEO done well: Google says there are no additional requirements to appear in AI Overviews or AI Mode and no special optimizations are necessary (Google Search Central). The practical work is narrower than the hype suggests: let the right crawlers in, write clear and well-sourced pages, keep structured data accurate, and measure results carefully.
This playbook is written for owners and marketing leads without an SEO team. It puts the steps in priority order and separates what the evidence supports from what vendors are selling.
SEO vs AEO vs GEO: what the terms mean
The three acronyms describe overlapping goals, not separate disciplines.
| Term | Goal | Where it shows up |
|---|---|---|
| SEO (search engine optimization) | Rank pages in traditional search results | Google and Bing blue links |
| AEO (answer engine optimization) | Be the direct answer to a question | Featured snippets, voice assistants, AI answers |
| GEO (generative engine optimization) | Be retrieved and cited inside AI-generated answers | Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Copilot |
The foundation is the same for all three. To be eligible for Google's AI features, a page must be indexed and eligible to show in Google Search with a snippet, and Google notes that indexing and serving are not guaranteed (Google Search Central). If your site has crawl, indexing or content quality problems, fix those before spending anything on "AI optimization."
The playbook in priority order
If you only have a few hours a month, work through these in order:
- Check crawler access so search-focused AI bots can reach your pages.
- Understand what the research actually supports so you don't waste time on tactics that didn't work.
- Rewrite your most important pages answer-first, with specific claims and sources.
- Make schema and business details match what is visible on your site and your profiles.
- Add llms.txt only if it's cheap, and expect little from it.
- Set realistic traffic expectations using the click studies.
- Measure with Search Console, analytics referrals and conversions.
Step 1: Check AI crawler access
This is the highest-value, lowest-effort step. Each AI company uses different crawlers for different jobs, and the key distinction is between search crawlers (which decide whether you can appear and be linked in answers) and training crawlers (which collect content to train models). You can allow one and block the other.
| Crawler | What it controls | Recommended setting for most SMBs |
|---|---|---|
| Googlebot | Google Search crawling and indexing, which AI Overviews and AI Mode depend on | Allow. Pages must return HTTP 200 and be indexable (Google Search Central blog) |
| Google-Extended | Use of your content for Gemini training and grounding. Google says it "does not impact a site's inclusion in Google Search" and is not a ranking signal (Google crawler docs) | Your choice. Blocking it does not affect Google Search, but it also opts your content out of Gemini grounding, not just training |
| OAI-SearchBot | Whether your site is surfaced in ChatGPT search. Sites that opt out "will not be shown in ChatGPT search answers," though they may still appear as navigational links (OpenAI bot docs) | Allow |
| GPTBot | Crawling for training OpenAI's foundation models (OpenAI bot docs) | Your choice. Independent of OAI-SearchBot |
| PerplexityBot | Surfacing and linking websites in Perplexity search results. Perplexity says it "is not used to crawl content for AI foundation models" (Perplexity bot docs) | Allow |
A robots.txt that allows search visibility while opting out of OpenAI model training looks like this:
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: GPTBot
Disallow: /
A few details matter:
- Blocking GPTBot does not remove you from ChatGPT search. OpenAI states the settings are independent; OAI-SearchBot controls search inclusion (OpenAI bot docs).
- Changes are not instant. OpenAI says robots.txt changes take about 24 hours to take effect (OpenAI bot docs), and Perplexity says up to 24 hours (Perplexity bot docs).
- User-triggered fetches are different. When a person asks ChatGPT or Perplexity to open a specific page, the fetch comes from ChatGPT-User or Perplexity-User. OpenAI says robots.txt rules may not apply to ChatGPT-User (OpenAI bot docs), and Perplexity says Perplexity-User "generally ignores robots.txt rules" (Perplexity bot docs).
- You can limit what Google quotes. Snippet controls such as nosnippet, data-nosnippet and max-snippet (or noindex) limit what Google shows from a page, including in its AI features (Google Search Central).
- Check more than robots.txt. Security plugins, CDN bot settings and hosting firewalls can block unfamiliar crawlers even when robots.txt allows them. Ask whoever manages your hosting to confirm.
Step 2: What the GEO paper actually found (and its limits)
Much of the GEO advice online traces back to one academic paper: "GEO: Generative Engine Optimization" by Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande, accepted to KDD 2024 (arXiv 2311.09735). The authors built a benchmark called GEO-bench with 10,000 queries and reported that GEO methods "can boost visibility by up to 40%" in generative engine responses, with effectiveness varying by domain.
Which tactics worked
The researchers tested nine content changes; seven of them are summarized below (GEO paper, full text):
| Tactic | Result in the study |
|---|---|
| Cite Sources, Quotation Addition, Statistics Addition | Best performers: 30-40% gains on Position-Adjusted Word Count, 15-30% on Subjective Impression |
| Fluency Optimization, Easy-to-Understand | 15-30% gains |
| Authoritative (more persuasive tone) | No significant improvement |
| Keyword stuffing | Little to no improvement; 10% worse than baseline on Perplexity.ai |
The pattern is useful: adding evidence (sources, quotes, numbers) helped, while sounding more confident or repeating keywords did not.
Who benefited, and who lost
The gains were uneven. Using the Cite Sources method, the source ranked fifth saw a 115.1% visibility increase, while the top-ranked source decreased 30.3%. The authors conclude that lower-ranked websites benefit significantly more from GEO (GEO paper, full text). That is encouraging for smaller sites, but it also means the effect partly reshuffles visibility among sources rather than adding it for everyone.
The limits
- It was a lab simulation. The main experiments used gpt-3.5-turbo over the top 5 Google results for each query, not live ChatGPT, Google AI Overviews or Perplexity in production (GEO paper, full text).
- It is a 2023-24 study. AI search engines have changed a great deal since then.
- "Up to 40%" is a ceiling, not an average. Results varied by domain and method.
Treat the paper as evidence about what kind of content gets used, not as a forecast for your site.
Step 3: Write answer-first, well-sourced pages
Microsoft's Bing team gives the clearest practical guidance here. Krishna Madhavan, a Principal Product Manager, wrote that AI assistants "don't read a page top to bottom like a person would" (Microsoft Advertising blog). Bing recommends self-contained, snippable answers, specific H2 and H3 headings, question-and-answer formats, lists and tables, and JSON-LD schema. It warns against hiding answers in tabs, putting core information in PDFs, leaving details only in images, and vague marketing claims.
Google's guidance points the same way: create unique, non-commodity content, give users a good page experience, and support text with images or video where it helps (Google Search Central blog). Google also explains that AI Overviews and AI Mode may use "query fan-out," running multiple related searches across subtopics and data sources (Google Search Central). Pages that cover a topic's real subquestions give those searches more to find.
A practical rewrite checklist for your top pages:
- Answer the main question in the first two or three sentences. State who you serve, what it costs or how it works, and where.
- Use specific headings. For example, "How much does a water heater replacement cost in Naperville?" beats "Our Pricing."
- Make each section stand on its own. A paragraph pulled out of context should still make sense.
- Cite sources for claims. Link to the regulation, study or manufacturer spec, matching the paper's strongest tactics.
- Replace adjectives with facts. For example, "Licensed in Illinois since 2012, 24/7 emergency service" is more citable than "industry-leading."
- Move key details out of PDFs, images and tabs into plain HTML text.
Your site's architecture affects how easily you can do this at scale. Our comparison of headless websites vs WordPress for lead generation covers the trade-offs for content-heavy sites.
Step 4: Schema and entity basics
Google is explicit that you do not need special schema.org markup to appear in its AI features (Google Search Central). What it does recommend is structured data that matches the visible content on the page, plus keeping your Merchant Center and Business Profile information current (Google Search Central blog).
For a typical small business, the basics are:
- Organization or LocalBusiness markup with the same name, address, phone and hours shown on the page.
- Service or Product markup only where the page actually describes that service or product.
- A complete, current Google Business Profile that matches your website.
- Consistent business details across your site, directories and social profiles, so AI systems can confidently connect mentions to one entity. This is practitioner guidance rather than a measured effect, but inconsistent details create avoidable ambiguity.
The rule of thumb: never mark up anything a visitor cannot see. FAQ markup for questions that aren't on the page, or review markup for reviews you don't display, creates risk without benefit.
Step 5: llms.txt, and why it is optional
llms.txt was proposed by Jeremy Howard on September 3, 2024. It is a Markdown file at /llms.txt with a site name as an H1 (the only required element), an optional blockquote summary and H2 sections of links (llmstxt.org). It is a proposal, not a standard.
Google's documentation says you do not need new machine-readable files or AI text files to appear in AI Overviews or AI Mode (Google Search Central). Google's John Mueller also commented on Reddit that "AFAIK none of the AI services have said they're using LLMs.TXT" and compared it to the keywords meta tag (Search Engine Journal). That was a comment, not official policy, but it matches Google's documentation.
If adding an llms.txt file is a quick task for your developer, there's little harm in it. Just don't treat it as a ranking lever or pay extra for it.
Step 6: What the click data says
The short answer to "will AI Overviews cut my traffic?" is that third-party data points toward fewer clicks, Google says the remaining clicks are better, and nobody has proven causation. Here are the main studies side by side.
| Source | Type | Data | Finding | Caveat |
|---|---|---|---|---|
| Pew Research Center (July 2025) | Independent research | Browsing data from 900 US adults, March 2025 | Users clicked a traditional result on 8% of visits with an AI summary vs 15% without; clicked a link inside the summary on 1% of visits | US adults only; AI Overview coverage has expanded since March 2025 |
| Ahrefs (Feb 2026) | Vendor study | 300,000 keywords, December 2023 vs December 2025, desktop Search Console data | An AI Overview "correlates with a 58% lower average clickthrough rate" for the position-1 page | Correlational; Ahrefs sells SEO software |
| Seer Interactive (Nov 2025) | Agency study | 3,119 terms (informational queries) across 42 client organizations, data through Sept 2025 | Q3 2025 organic CTR: 1.45% with no AI Overview, 0.70% with an AI Overview citing the brand, 0.52% with one not citing it | Seer says it "cannot definitively prove" citation causes the higher CTR; Seer sells SEO services |
| Google (May 2025) | Platform claim | Not published | Clicks from results pages with AI Overviews are "higher quality," with users more likely to spend more time on site | No supporting data released; conflicts with the third-party CTR studies |
Pew also found that users ended their browsing session on 26% of pages with an AI summary vs 16% without, that about 18% of 68,879 searches produced an AI summary, and that Wikipedia, YouTube and Reddit made up 15% of AI summary sources (Pew Research Center).
What this means for planning: informational, top-of-funnel content is the most exposed. Our reading of these studies: pages that serve people ready to act (service pages, pricing, local pages, booking) are less exposed and deserve more of your effort. It is also a strong argument for owning your audience relationships, which we cover in owned marketing stacks vs SaaS tools.
Step 7: Measuring AI visibility
There is no single "AI rankings" report, so combine three sources and focus on business outcomes.
Search Console
Google counts traffic from AI Overviews and AI Mode in Search Console's Performance report under the "Web" search type. There is no separate AI report (Google Search Central). Watch impressions and clicks on your most important queries over time, and note where impressions rise while clicks fall.
Referral traffic
In your analytics tool, look at referral visits from AI assistants such as chatgpt.com and perplexity.ai. You may see tagged URLs from some tools, but don't build your reporting on any parameter the platform hasn't documented. Treat referrals as a floor, since many AI answers produce no click at all.
Prompt sampling
Write 20 to 50 questions your customers actually ask, run them through ChatGPT, Perplexity, Google AI Mode and Copilot on a regular schedule, and record whether you're mentioned or cited. Paid tracking tools automate this. Either way, results vary by phrasing, location and model, so treat the output as directional, not as a ranking.
Conversions
Google itself recommends measuring conversions, not clicks alone (Google Search Central blog). Ask new leads how they found you, and add "ChatGPT or another AI tool" as an option on intake forms. If you need help connecting search, content and pipeline into one plan, that is the territory of growth strategy consulting.
Limitations: what GEO cannot do
Be skeptical of anyone who sells GEO as a guaranteed outcome.
- Nobody can guarantee citation. Allowing crawlers and writing clear pages only makes your site eligible. Google states indexing and serving are not guaranteed, and AI answers change with each prompt.
- The research base is thin. The best-known study is a 2023-24 lab simulation on gpt-3.5-turbo. Live engines work differently and keep changing.
- Click studies are correlational. Ahrefs sells SEO software and Seer sells SEO services, Pew measured one month of US browsing in early 2025, and Google's "higher quality clicks" claim has no published data.
- Visibility tracking is noisy. Prompt-based tools sample a small slice of possible questions.
- It won't fix fundamentals. If your site isn't indexed, loads slowly, or says nothing a competitor doesn't, AI-specific tactics will not compensate.
- It may not be your best channel. If most of your customers come from referrals, local maps or paid search, a few hours on your Business Profile and service pages may beat a broad GEO program.
GEO in one paragraph: what to do this month
GEO is mostly good SEO with a few new decisions. Allow the search crawlers (Googlebot, OAI-SearchBot, PerplexityBot), decide separately whether to allow training crawlers, write specific answer-first pages backed by sources, keep structured data and business profiles accurate, and measure leads rather than chasing AI mentions. Skip anything sold as a shortcut, including persuasive tone, keyword stuffing and paid llms.txt packages.
If you'd like a second set of eyes on your crawler settings, content and measurement, Robustech offers SEO, AEO and GEO services built on the same evidence-first approach.
