Article Overview – PDF Here –
This article examines what Google’s 2026 guidance for generative AI features in Search actually asks of small businesses and solopreneurs. It separates documented Google Search requirements from broader GEO and AEO claims, reviews Google’s preference for non-commodity content, tests popular tactics such as llms.txt, chunking, mention chasing, AI-specific rewrites, and special schema, and compares those claims with current external research. It then applies Factics to a practical weekly workflow and explains what Search Console can and cannot measure, including dedicated generative impressions without a dedicated generative click metric.
Vendors still package Generative Engine Optimization as a separate layer of work, often bundling llms.txt, schema, entity work, and multi-engine citation tracking into paid services. One current agency page advertises a ten-step GEO playbook that includes schema, directories, listicles, /llms.txt, and weekly citation tracking across several AI systems (DoodleWeb, 2026). That documents the sales pitch in current practice. It does not establish that those tactics improve visibility in Google Search.
Google published its generative AI Search guidance on May 15, 2026, and last updated the guide on July 10. The message is quieter than the sales pitch. Generative AI features on Search remain rooted in Google’s core ranking and quality systems, so foundational SEO still applies. Google also says unique, compelling, useful content will likely influence a site’s presence in generative AI search in the long run more than any other suggestion in the guide (Google Search Central, 2026a; Mueller, 2026).
That lands where I have practiced digital strategy since 2009, and where Factics begins: every fact pairs with a tactic, and every tactic needs a KPI defined before the work starts (Puglisi, n.d.).
Is SEO still relevant for AI Overviews and AI Mode?
Yes. Google states that SEO best practices remain relevant because its generative AI features rely on the same core Search ranking and quality systems used to retrieve pages from the index (Google Search Central, 2026a).
At a high level, two techniques matter for how those features work. Retrieval-augmented generation, also called grounding, uses core ranking systems to retrieve relevant and current pages, then reviews specific information from those pages to support a response with clickable links. Query fan-out generates concurrent related queries so the system can fetch additional results that address the original ask. Google frames both as extensions of Search rather than a separate optimization universe.
“AEO” and “GEO” appear in the market as labels for visibility work aimed at AI search experiences. From Google Search’s perspective, optimizing for generative AI search is still optimizing for the search experience, and thus still SEO. Google tells site owners to evaluate third-party AEO and GEO advice against official guidance rather than treating a new label as proof of a new ranking system (Google Search Central, 2026a).
The SMB stake is not that Google uses a different ranking system for small businesses. The stake is opportunity cost. An enterprise can test optional tactics with a separate team. A solopreneur or small business has the same finite week for client work, operations, publishing, and measurement. Time spent maintaining a Search-specific tactic that Google says is unnecessary is time not spent creating work only that business could credibly publish.
Google’s own guide adds an SMB-relevant point that deserves more attention. Local businesses and ecommerce operators can use Google Business Profiles and Merchant Center to help products and services appear in AI responses and other Search results (Google Search Central, 2026a). That is ordinary platform hygiene tied directly to Google’s documented system, not a speculative GEO layer.
What does Google mean by non-commodity versus commodity content?
Google draws a contrast that any working marketer can test against a draft queue. Commodity content often rests on common knowledge anyone could assemble and typically adds little unique insight. Google’s example is a piece like “7 Tips for First-Time Homebuyers.” Its non-commodity counterpart is “Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line” (Google Search Central, 2026a).
The distinction is not that personal storytelling automatically ranks. It is that first-hand experience can supply information that a generic summary cannot. A review grounded in actual use, a client case with evidence, an operational failure, a field observation, or a specific decision made under real conditions can add information that is not already replicated across the web.
Organization still matters. Google’s guide recommends paragraphs, sections, headings, useful images, and video where they support the reader. It also warns against manufacturing a separate page for every query variation primarily to manipulate rankings or generative responses. That pattern can violate the scaled content abuse policy and is not a durable quality strategy (Google Search Central, 2026a).
Google’s companion people-first guidance was updated on October 1, 2026. It says content should primarily help people rather than manipulate rankings, and it states that trust is the most important element among the E-E-A-T concepts. It also keeps the Who, How, and Why framing for authorship, production process, and purpose (Google Search Central, 2026b).
The same October 1 update cycle strengthened the human review point. Google’s generative AI content guidance now says manual fact-checking and review of AI-generated content are critical before publication, including review of metadata, structured data, and image alt text (Google Search Central, 2026c). Google suggests adding information about how automated content was created when that context would help the audience. My publishing rule remains stricter: AI-assisted work carries #AIassisted disclosure and clears a human edit gate before publication.
That is the non-commodity marker I defend here. Factics does not turn experience into a ranking factor. It turns a claim into a governed decision: what is known, what action follows, and what evidence will show whether the action worked.
Which popular GEO and AEO tactics can SMBs ignore for Google Search?
Google collected circulating myths and named what site owners can ignore for Google Search.
llms.txt files and similar machine-readable files are not required. Google Search says it does not use them for its generative AI capabilities. Creating them for other services neither helps nor harms visibility or rankings in Google Search because Search ignores them (Google Search Central, 2026a).
That Google-specific boundary matters. An llms.txt file may still have uses in other technical or agentic contexts. The claim here is narrower: it is not a Google Search visibility requirement. Independent server-log evidence points in the same direction. A June 2026 Ahrefs study covering 137,210 domains found that 28 percent published an llms.txt file, while 97 percent of valid files received no requests during the study month (Linehan, 2026a). That does not establish that the file has no use elsewhere. It does make a strong case against treating it as a required AI-search deliverable.
Content chunking is not a requirement either. Google says its systems can understand multiple topics on a page and surface the relevant portion. There is no ideal page length, so the page should fit the audience and subject rather than an imagined AI parser.
Rewriting content just for AI systems is unnecessary for Google Search. Google’s systems understand synonyms and general meanings, so publishers do not need to capture every long-tail variation through repeated rewrites.
Seeking inauthentic mentions is not as helpful as it may seem. Google’s generative features can surface what the web says about products and services, but the guide distinguishes that from manufacturing mentions merely to influence visibility. Earned references and manipulative mention chasing are not the same activity.
Structured data is also more nuanced than the sales pitch. Google says structured data is not required for generative AI search, and no special schema.org markup unlocks those features. Structured data remains useful for ordinary SEO where it supports rich-result eligibility (Google Search Central, 2026a). An Ahrefs study tracked 1,885 pages that added JSON-LD schema and found no major citation uplift across Google AI Overviews, AI Mode, or ChatGPT after matching against controls (Linehan, 2026b). That is industry evidence rather than a Google ranking disclosure, but it helps separate correlation from a causal claim.
Fact: Google says site owners can ignore llms.txt, chunking as a requirement, AI-only rewriting, inauthentic mention chasing, and special structured-data schemes for generative AI visibility in Google Search.
Tactic: Cap those Search-specific experiments at one hour per week and redirect the recovered time to source-backed original observation.
KPI: Record the baseline first. Then hold hack-task time at or below one hour per week and original-draft time at or above four hours per week for eight weeks. If the time allocation does not change, the tactic did not land.
Does GEO research contradict this argument?
No, but it prevents an overbroad conclusion.
The foundational GEO paper by Aggarwal and colleagues reported visibility gains of up to 40 percent in its experimental generative-engine setting. That work matters because it shows that content treatment can affect visibility in generative responses under defined conditions (Aggarwal et al., 2024). It does not establish that llms.txt, chunking, synthetic mentions, or special schema improve organic discoverability in production Google Search.
A 2026 critical survey by Martinez reviewed 45 studies and drew the distinction more sharply. The review found evidence that already-retrieved content can sometimes be changed in ways that affect citation or use, but it found no stable, longitudinal, cross-platform causal technique for organic discoverability or downstream behavior across the reviewed literature (Martinez, 2026).
That is the boundary this article needs: GEO is not imaginary, and generative systems can respond differently to how content is written and presented. What does not follow is that every tactic sold under the GEO label creates a durable Google Search advantage. Google’s own Search-specific guidance rejects several of the tactics now sold as requirements.
What weekly Factics workflow replaces hack chasing?
Factics asks three questions of any recommendation: what do we know, what do we do with what we know, and what happened because we acted. A KPI written after results arrive becomes a story about the past. A KPI written first is a commitment that can fail.
Before week one, confirm that the property is included in Search generative AI features in Search Console. Google now provides a property-level control for inclusion in AI Overviews, AI Mode, and generative features in Discover. Inclusion is the default, and exclusion removes the site’s links and content from those features without becoming a ranking signal for the rest of Search (Google Search Console Help, 2026b).
Then apply the loop.
Fact: Google says unique, compelling, useful content will likely influence generative AI presence in the long run more than any other suggestion in its guide.
Tactic: Publish one experience-backed piece per week instead of filling the calendar with commodity listicles.
KPI: For eight weeks, track Generative AI performance report impressions per URL for the experience-backed set. Compare those pages with a matched set of commodity pages from the same site under similar topic, age, and publishing conditions. Track standard Search clicks for the same URLs as a separate signal, but do not label those clicks as dedicated Generative AI report clicks. Also track the share of new posts carrying a clear first-hand marker near the opening, with a target of at least 80 percent. That 80 percent target is an internal Factics measure, not a Google ranking factor.
Fact: Google’s October 1 guidance says AI-generated content requires manual fact-checking and review before publication.
Tactic: Keep the human edit gate and #AIassisted disclosure on AI-assisted work.
KPI: 100 percent of AI-assisted posts clear the human gate and carry disclosure before publication. Engagement can be compared with the site’s normal content baseline, but that comparison should be treated as operational evidence rather than a causal finding about disclosure.
A workable week stays concrete. Monday, choose one observation from client work, product use, or a failed experiment that can be defended. Tuesday, draft against verified sources and reject invented citations; Wednesday, conduct the human review and attach disclosure when AI assisted. Thursday, publish one URL instead of a swarm of near-duplicates. Friday, log the predefined KPIs and leave the hack backlog untouched unless a task still earns its one-hour cap.
That cadence is custody. AI-assisted content with verified sources and a human checkpoint is publishable. A pile of near-duplicate variations created mainly to manipulate fan-out or rankings is not a strategy Google recommends, and it can cross into scaled content abuse.
How do you measure presence in generative AI features without vanity metrics?
Google’s dedicated Generative AI performance report now covers AI Overviews and AI Mode in Search. It rolled out worldwide on August 31, 2026, and exposes impressions by page, country, date, and device (Google Search Console Help, 2026a).
The dedicated report does not expose clicks, CTR, query data, or average position. That is the correction this article needed most.
Clicks from AI Mode and AI Overviews still count in Search Console’s broader Performance methodology when a user clicks an external link. Those clicks remain part of the wider Search reporting rather than a dedicated click metric inside the Generative AI performance report (Google Search Console Help, n.d.). The distinction matters because the two data sources answer different questions.
The Generative AI report answers: Was this URL shown in Google’s generative Search features?
The broader Search Performance report answers: Did this URL receive Search clicks and impressions overall?
It does not let the publisher cleanly attribute a specific click total to the dedicated Generative AI report.
Vanity looks like screenshot theater: one AI Overview citation with no URL-family context, or a vendor dashboard presented as if it were Google’s internal ranking data. Google explicitly warns that third-party tools do not have access to its internal ranking or AI systems (Google Search Central, 2026a).
Measurement that survives a quarter is less dramatic. Track dedicated generative impressions per URL, standard Search clicks as a separate non-attributed signal, time moved from hack tasks to original drafts, the share of posts with first-hand evidence, and disclosure compliance.
The falsification condition should be equally plain. If, over a full quarter in this niche, a matched set of commodity pages produces stronger per-URL Generative AI impressions than the experience-backed set under comparable topic and page-age conditions, the non-commodity priority weakens for that site. The result would still be observational rather than causal, but it would be evidence against the current operating assumption.
What changes if you stop optimizing for theater?
The structural consequence is editorial, not technical. When Google says optimizing for generative AI search is still SEO, the durable play for solopreneurs and SMBs is non-commodity, people-first work supported by ordinary technical SEO, Google Business Profile or Merchant Center where relevant, and a Factics loop that defines the measure before publication.
Search-specific hack stacks that Google says to ignore become optional experiments, not the calendar.
That conclusion is scoped to Google Search. ChatGPT, Perplexity, Claude, Copilot, Gemini outside Search, and other answer systems have different retrieval and citation behaviors. Any tactic aimed at those systems should be treated as a separate experiment with its own evidence and KPI rather than imported from Google’s rules or projected back onto them.
Watch the Generative AI performance report as more experience-backed URLs accumulate enough data to compare. Watch whether Google’s ignore list changes. Watch whether new independent research produces stable causal evidence for one of the tactics currently sold as a requirement. Until then, spend the week on the piece only you could write.
References
Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., & Deshpande, A. (2024). GEO: Generative Engine Optimization. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. https://arxiv.org/abs/2311.09735
DoodleWeb. (2026, September 22). GEO agency: Generative engine optimization for AI search. https://doodleweb.com/geo-agency
Google Search Central. (2026a, July 10). Optimizing your website for generative AI features on Google Search. Google for Developers. https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
Google Search Central. (2026b, October 1). Creating helpful, reliable, people-first content. Google for Developers. https://developers.google.com/search/docs/fundamentals/creating-helpful-content
Google Search Central. (2026c, October 1). Google Search’s guidance on generative AI content on your website. Google for Developers. https://developers.google.com/search/docs/fundamentals/using-gen-ai-content
Google Search Console Help. (2026a, August 31). Generative AI performance report (Search). https://support.google.com/webmasters/answer/16984139?hl=en
Google Search Console Help. (2026b, August 31). Search generative AI control. https://support.google.com/webmasters/answer/16908024?hl=en
Google Search Console Help. (n.d.). What are impressions, position, and clicks? https://support.google.com/webmasters/answer/7042828?hl=en
Linehan, L. (2026a, June 15). We analyzed 137K sites: 97% of llms.txt files never get read. Ahrefs. https://ahrefs.com/blog/llmstxt-study/
Linehan, L. (2026b, May 11). We tracked 1,885 pages adding schema. AI citations barely moved. Ahrefs. https://ahrefs.com/blog/schema-ai-citations/
Martinez, O. (2026, July 15). Optimizing visibility in generative engines: A critical survey of Generative Engine Optimization (2023–2026). arXiv. https://arxiv.org/abs/2607.14035
Mueller, J. (2026, May 15). A new resource for optimizing for generative AI in Google Search. Google Search Central Blog. https://developers.google.com/search/blog/2026/05/a-new-resource-for-optimizing
Puglisi, B. C. (n.d.). Factics: The method behind valuable content and trusted AI. BasilPuglisi.com. https://basilpuglisi.com/factics/
What does GEO mean for Google Search?
GEO means generative engine optimization, a label used for work aimed at improving visibility in AI search experiences. Google says that for its own Search product, generative AI features remain rooted in core Search ranking and quality systems, so optimizing for those features is still SEO rather than a separate ranking system.
Does Google Search use llms.txt files?
No. Google says Search does not use llms.txt or similar special machine-readable files for its generative AI capabilities. Creating one for another service does not help or harm visibility in Google Search. Independent server-log research also found that most valid llms.txt files received no requests during the measured month.
Does structured data improve generative AI visibility in Google Search?
Google says structured data is not required for generative AI Search and that no special schema.org markup unlocks AI Overviews or AI Mode. Structured data still serves ordinary SEO purposes such as rich-result eligibility. An Ahrefs study found no major citation lift after matched pages added JSON-LD schema.
What should an SMB prioritize instead of GEO hacks?
The article recommends prioritizing foundational SEO, crawlable pages, useful local or ecommerce data where relevant, and original non-commodity content grounded in real experience. For a small business, the issue is opportunity cost: time spent on Search-specific tactics Google says are unnecessary is time not spent producing evidence-backed work customers can use.
What can Search Console measure for Google generative AI features?
Google’s dedicated Generative AI performance report shows impressions for AI Overviews and AI Mode, broken down by page, country, date, and device. It does not expose dedicated clicks, CTR, query data, or average position. Broader Search Performance reporting still counts external-link clicks, but those clicks are not isolated as a generative-only metric.
Does GEO research show that optimization never works?
No. Research shows that content treatment can affect visibility or citation in some generative-engine experiments after content is retrieved. The evidence does not establish a stable, cross-platform technique for organic discoverability in production Google Search. That distinction is why the article keeps broader GEO research separate from Google’s Search-specific guidance.
The Other AI: Audio Briefings on Augmented Intelligence and AI Governance
Spotify | Apple Podcasts | Amazon Music | YouTube Playlist
#AIassisted using the HAIA Ecosystem | CC BY-NC-SA 4.0 Free for personal, educational, and noncommercial research use with attribution. Commercial exploitation, paid productization, and enterprise commercialization require separate permission and licensing.
Leave a Reply