llms.txt for a Qatari Site: A Working Template

llms.txt does not change search ranking or search visibility for a Qatari site in any way. Google Search Central states that Google Search does not use llms.txt or similar “AI text” files for generative AI features on Search, and its changelog clarifies that maintaining such files will not positively or negatively affect visibility or rankings. Search crawlers do not treat the file as a ranking signal; Google largely ignores it for Search. For Doha and wider Qatar publishers, the file is an optional Markdown map that some LLM tooling may read after it already knows the site exists. It is not SEO, not GEO magic, and not a substitute for bilingual content quality, entities, schema, or a correct robots.txt policy.

What llms.txt actually is

Jeremy Howard’s proposal at llmstxt.org defines `/llms.txt` as a Markdown file that helps agents use a website after they already care about it. The original Answer.AI post (3 September 2024) framed the problem as context assembly: HTML is noisy, context windows are finite, and site authors know which pages matter for documentation or organisational facts. Version 2 keeps that purpose and adds link-relation hints so agents can find Markdown mirrors of pages. The specification never promised discovery by Googlebot, Bingbot, or “AI search ranking.”

The required structure is small. An H1 names the project or company. An optional blockquote summarises the entity. Unheaded paragraphs may add guidance. H2 sections then list curated links as Markdown list items with optional notes. An “Optional” H2 is convention for secondary links that tools may skip when context is tight. That is the whole format. There is no ranking weight, no crawl budget rule, and no guarantee that ChatGPT, Claude, Perplexity, Gemini, or Bing will fetch the file during ordinary retrieval.

For a company on a `.qa` domain in West Bay or Lusail, the honest use case looks like this: a coding agent or a publisher-chosen tool loads the file to learn legal name, primary EN/AR URLs, contact paths, and policy pages. The dishonest use case circulating on LinkedIn treats the same file as a ranking patch for “AI Overviews” or “GEO.” Those claims contradict both the proposal’s stated goal and Google’s published Search guidance.

Google’s official stance: no ranking effect

Google’s public line leaves no room for ranking fantasies on Qatari domains or anywhere else. The Search Central guide Optimizing your website for generative AI features on Google Search places LLMS.txt under mythbusting: site owners do not need new machine-readable AI text files, markup, or Markdown to appear in Google Search, including generative AI capabilities, because Google Search itself does not use them. The same page tells readers to prioritise ordinary SEO over “AEO/GEO hacks,” and names creating unnecessary AI text files such as llms.txt among tactics to ignore for Google Search. That language is about Search visibility in Doha SERPs and AI Overviews, not about whether a WordPress plugin can emit a Markdown brochure for optional tooling. Treat every contrary agency claim as marketing until it cites Search Central.

What Google Search Central documents say

The AI optimisation guide is the primary citable document. It groups llms.txt with other special markup that Google Search does not use for generative features. It separately notes that Google may crawl and index many non-HTML files without treating them as special ranking formats. In plain language: even if Googlebot happens to fetch `/llms.txt`, that fetch is not a ranking API and does not unlock AI Mode preference. The guide’s “next steps” section repeats the instruction to ignore unnecessary AI text files for Google Search while investing in crawlable, people-first content and sound technical structure.

Qatar site owners should treat that guide as authoritative for Google surfaces used heavily in Doha: classic results, AI Overviews, and AI Mode. Local campaigns that sell “llms.txt installation for Google AI ranking” are selling a service Google has already filed under myths. Budget that hour instead on Arabic page quality, correct `Organization` schema with a Qatar address, and Search Console coverage of the EN/AR URL pairs that actually earn impressions. llms.txt remains irrelevant to those ranking and eligibility paths.

What Google’s changelog and clarifying note add

Google addressed community confusion in its Search Central updates log. The changelog entry on clarifying llms.txt states that the files are not needed for Google Search and will not negatively or positively impact visibility or rankings, while remaining fine to maintain for other services that actually consume them. Industry reporting from Search Engine Roundtable and Search Engine Journal quoted the clarifying note that creating and maintaining LLMS.txt files will not harm nor help visibility or rankings in Google Search, because Google Search ignores them. Those sources matter for Qatar marketing teams that saw June 2026 LinkedIn threads claiming Google “approved” the file as an AI SEO tactic.

Approval to keep a harmless optional file is not approval of a ranking lever. Early phrasing that special files were unnecessary for generative AI search was sometimes quoted as a blanket ban on publishing Markdown helpers for any AI system. Google narrowed the claim to Google Search and added that other systems may still use such files. The practical reading for Qatar publishers is dual: optional for third-party tooling; inert for Google ranking and Google generative visibility. Some teams delayed shipping documentation Markdown because they worried Google would penalise “AI SEO files.” Google’s published position does not describe a penalty. Harm and help are both denied. The opportunity cost remains: hours spent polishing a vanity `/llms.txt` while Arabic service pages stay thin.

Why Mueller’s comments matter for Qatar marketers

John Mueller’s comments close the endorsement loophole. When someone pointed at `ai.google.dev` llms.txt files and asked whether that was Google endorsement, Mueller answered on Bluesky that it was not. On Search Off the Record coverage reported by Search Engine Journal, Mueller said he spoke with someone involved in the proposal: the idea was never to make discovery by search or LLM systems easier; it was closer to helping a system that already knows the site find what else is there. He also argued that self-reported “best pages” files cannot differentiate competing sites for discovery, because every publisher can claim primacy.

For a Doha clinic, Lusail developer, or QFC-registered consultancy, that means `/llms.txt` will not decide who Google cites in AI Overviews. Indexed HTML quality still does. Mueller’s Bluesky “no” closes the “Google uses it on its own sites, so it must be a ranking factor” argument. Different Google properties can ship documentation helpers for coding agents without the Search ranking team adopting those helpers as signals. His discovery critique matches what Qatar agencies see in client decks: a file written by the seller claiming the seller is the best source. Retrieval systems that ground answers in the open web still need independent differentiation: links, brand mentions, entity consistency, and page usefulness. Publish `/llms.txt` only if a concrete tool in the stack will read it. Do not put it on a scorecard next to Core Web Vitals or hreflang health. Do not promise leadership teams that the file moves Google visibility in Qatar SERPs. When an agency pitch slides from “optional agent map” into “boosts AI search,” demand a citation to Google Search Central. The available citations say the opposite.

robots.txt versus llms.txt

Confusion between these two root files drives most of the hype sold to Qatar marketing teams. Both sit at well-known paths. Only one is part of a formal crawl-control protocol that major search engines actually honour before fetching. The Robots Exclusion Protocol (RFC 9309) and Google’s robots.txt interpretation docs describe how crawlers download rules and decide allow or disallow paths. llms.txt, by the proposal author’s own framing on llmstxt.org, is for on-demand context when an agent already needs information about a topic. It does not grant or deny crawl permission. A Doha publisher that fixes robots.txt can change whether Googlebot or OAI-SearchBot may fetch `/ar/services/`; adding llms.txt does not change those permissions at all. Keep the two files mentally separate in every sprint checklist.

What robots.txt controls for AI and search bots

Google’s common-crawler documentation lists Googlebot as the agent whose preferences affect Google Search features, and it documents `Google-Extended` as a separate product token that does not impact inclusion in Google Search and is not used as a ranking signal. OpenAI publishes distinct bots for training (`GPTBot`), search indexing (`OAI-SearchBot`), and user-triggered fetches (`ChatGPT-User`). Anthropic documents Claude-related crawlers that honour robots.txt. These are the levers that actually change whether a Qatar page can enter an AI search index or a training corpus.

A practical robots.txt for a public marketing site on `example.qa` usually allows Googlebot and Bingbot, decides explicitly on search-oriented AI bots the business wants citations from, and optionally disallows training-oriented agents after legal review. None of those lines belong in llms.txt. Mixing them teaches staff the wrong mental model: that a Markdown list is a crawl policy.

What llms.txt never controls

llms.txt cannot unblock a Disallow. It cannot force Google to prefer the Arabic URL. It cannot override `noindex`. It cannot repair soft-404 service pages in Industrial Area microsites. It cannot replace a missing Google Business Profile for a clinic in Al Waab. Treating it as a control plane wastes the only scarce resource most Qatar SMEs have: engineering hours from a small web team.

The proposal itself contrasts the file with robots.txt and sitemap.xml. robots.txt states acceptable access. sitemap.xml lists indexable human pages at scale. llms.txt offers a curated overview sized for context windows, often pointing to Markdown versions of key docs. That contrast is educational, not promotional. Curated overviews help agents that choose to load them. They do not rewrite Search ranking systems.

How Qatar teams should assign ownership

Give robots.txt and crawler policy to whoever owns security, legal, and SEO together. Give sitemap generation to the CMS or deploy pipeline. Give llms.txt, if used, to whoever owns documentation accuracy: usually marketing ops plus a developer who can keep absolute URLs correct after a migration from `www` to apex or from `/en` path prefixes to subfolders. Review the Markdown file when the commercial registration name, phone, or flagship service URLs change. Do not schedule weekly “AI SEO llms.txt optimisation” retainers. There is nothing weekly to optimise for ranking because ranking is not in scope.

If training opt-outs matter under internal policy or Qatar Personal Data Privacy Protection Law discussions with counsel, those decisions live in robots.txt user-agent groups and in product-specific tokens such as `Google-Extended`, not in a Markdown brochure. The comparison table below summarises how Qatar teams should prioritise each root file and adjacent signal. llms.txt stays optional and non-ranking; robots.txt and bilingual HTML stay critical.

Table. Root files and signals Qatari sites confuse with ranking levers

The LinkedIn and influencer hype problem

A familiar post pattern sells certainty: add `/llms.txt`, get cited by ChatGPT, climb AI Overviews, “future-proof GEO.” The post rarely links Google’s AI optimisation guide. It rarely quotes Mueller. It rarely admits that major systems may never fetch the file during ordinary answer generation. Adoption counts (“thousands of sites”) get framed as proof of ranking impact. Adoption proves copy-paste behaviour, not retrieval causality.

Google’s mythbusting section exists because this marketplace of tactics grew faster than evidence. The guide groups unnecessary AI text files with content chunking rituals, AI-only rewrites, and inauthentic mention campaigns. For Qatar, the commercial damage is concrete. A hospitality group in The Pearl spends budget on a generator plugin while its Arabic menu pages remain machine-translated mush. A contractor in Mesaieed ships llms.txt the same week its case studies stay PDF-only and uncrawlable. The file becomes a status badge in a slide deck.

Debunking does not require anger. It requires a burden of proof. If a vendor claims llms.txt boosts AI search for a `.qa` domain, ask which system reads it on a schedule, which log lines prove the fetch, and which controlled test showed citation lift after isolating other changes. Google’s documents already answer the Google half of that claim: Search ignores the file for ranking and generative visibility. For other answer engines, absence of public ranking specs plus Mueller’s differentiation argument still make “boost” claims speculative. Optional infrastructure is allowed. Speculative ranking claims are not.

Chrome Lighthouse agentic-browsing checks that notice a missing llms.txt also get mis-sold as SEO audits. An optional agent readiness hint is not a Search quality score. Passing or failing that check does not move Doha local pack rankings. Teams that report Lighthouse SEO scores to boards should keep agentic checks in a separate column labelled experimental tooling, not organic growth.

A working template for a Qatari bilingual site

A workable Qatar file follows llmstxt.org ordering and fits a bilingual `.qa` company: English and Arabic public pages, a Commercial Registration reference, Doha contact details, and curated links an agent might need after a user already asked about the firm. Absolute HTTPS URLs matter because relative links break when a tool fetches the file outside the site origin. Keep the H1 as the public brand name people use in Doha. Put the legal entity name in the summary if it differs from the brand. State plainly that the file is informational, that HTML pages hold canonical facts, and that the file does not affect search rankings. That disclaimer will not change Google’s behaviour; it reduces the chance a careless tool treats Markdown as fresher than the live site.

Entity fields that belong in the summary

The blockquote should carry identity, not slogans. Useful fields for Qatar: brand name, legal name if different, city (Doha / Lusail / Al Rayyan), CR number as published on the site, primary phone in +974 format, and the two homepage URLs. Avoid inventing Wikidata QIDs or “knowledge graph IDs” inside the file. If the organisation has a Wikidata item, link the public page under Optional only when the URL is stable. Entity building still happens on the open web and in structured data, not through a private boast list.

Do not paste client secrets, staging URLs, or intranet docs. llms.txt is public. Anything linked is an invitation to fetch. For regulated verticals (healthcare, banking under QCB expectations, legal services), keep links to pages already cleared for public marketing. Repeat the no-ranking sentence in the summary so sales decks cannot quietly delete the limitation when they paste the file into a proposal.

Bilingual EN/AR linking without splitting the story

List English and Arabic counterparts as separate bullets with clear labels. Agents and humans both benefit from explicit language tags in the link title. Do not rely on a single URL with a language switcher and hope the agent “figures it out.” If the site uses `example.qa` for English and `example.qa/ar/` for Arabic, say so once in the summary so tools do not treat `/ar/` as a different organisation.

hreflang remains the search-facing mechanism for locale selection. llms.txt is not a substitute for `hreflang` annotations or for equal content depth. A thin Arabic stub linked from a rich English llms.txt entry still leaves Arabic users and Arabic-language retrieval under-served. The file can expose the gap; it cannot fill it. Many Qatar teams keep the root llms.txt mostly English for parser simplicity and point to Arabic HTML pages as first-class links; that is acceptable and still does not create ranking upside.

Key pages worth listing for most .qa companies

Prioritise About, Contact, flagship Services, Privacy, and any regulated disclosures the firm already publishes. Add pricing only if the HTML page is genuinely maintained. Add case studies when they are HTML, indexed, and specific to Qatar work (Hamad Port logistics software, Lusail property portals, education platforms in Education City), not generic stock narratives. Skip mega-menus of every blog tag. The proposal rewards curation. A 400-link dump recreates the sitemap problem the format tried to avoid: too large for useful context.

If the company runs public API or developer docs, that is the strongest historical fit for llms.txt. Point agents at Markdown API references first. Marketing brochure sites gain less, which matches Mueller’s caution that Markdown shoe-spec mirrors are not a sales strategy. Ship the finished file at `https://www.example.qa/llms.txt` with UTF-8 encoding. Replace bracketed fields in the copy-paste example below. Section headings in a live file use Markdown H2 markers; the sample uses `H2:` labels only so this article’s outline stays clean.

“`text H1: Acme Digital Qatar > Acme Digital Qatar is a Doha-based web and automation agency registered in Qatar (CR: [CR-NUMBER]). Primary public site: https://www.acme.qa (English) and https://www.acme.qa/ar/ (Arabic). Contact: +974 [NUMBER], [area], Doha. This file is an optional machine-readable map for tools that choose to read llms.txt. It does not affect search engine rankings.

Acme publishes bilingual service and policy pages. Prefer the HTML URLs below as canonical. Markdown mirrors, if present, end with .md and are summaries for agents, not replacements for the live pages. For crawl permissions, see https://www.acme.qa/robots.txt. For URL discovery by search engines, see https://www.acme.qa/sitemap.xml.

H2: Company – About (English): Ownership summary, Doha office, leadership – About (Arabic): Arabic entity description for local readers – Contact: Form, phone +974, map pin for West Bay / selected area – Commercial details: Legal name, CR reference, licensing notes as published on-site

H2: Services – Web development: Public service scope in English – Web development (Arabic): Arabic service page – Business automation: Automation offers for Qatar SMEs – SEO and AI visibility consulting: Consulting scope; not a ranking guarantee page

H2: Policies – Privacy policy: Personal data notice for site visitors – Privacy policy (Arabic): Arabic privacy page – Terms: Website terms of use

H2: Optional – Blog index: English articles – Careers: Hiring page if active – Google Business Profile: Public GBP link if stable “`

When publishing the real file, convert each `H1:` / `H2:` line to a Markdown `#` / `##` heading per llmstxt.org. Keep the blockquote and the explicit ranking disclaimer. Re-check every URL after deploy. A broken Company link teaches agents the wrong office facts faster than having no file at all.

What actually matters for AI citation in Qatar

Retrieval systems that cite websites still depend on discoverable, trustworthy pages. For Google’s generative features, Search Central’s own guide returns readers to foundational SEO: unique non-commodity content, crawlable technical structure, solid page experience, and appropriate local or product feeds. Structured data is not required for generative AI search on Google, yet remains useful for rich results eligibility. That nuance matters: schema helps classic Search features; it is not a secret llms.txt replacement either.

Qatar-specific citation hygiene looks ordinary and hard. Publish accurate Arabic and English pages that name districts, regulators, and service constraints correctly. Keep `Organization` or `LocalBusiness` JSON-LD aligned with the Google Business Profile and the Commercial Registration public identity. Earn mentions on real Qatar industry sites, chamber pages, and news outlets instead of buying fake “AI mention” packages. Allow the search-oriented crawlers the business wants citations from, using robots.txt, after counsel reviews training opt-outs. Host HTML that renders primary content without hostage JavaScript. Fix soft 404s on Arabic URLs.

llms.txt sits outside that list. It may help a documentation agent that a developer points at the domain. It may help internal RAG tooling that a Qatar enterprise configures to prefer the file. Those are product choices. They are not organic search ranking. Teams that collapse “agent convenience” into “GEO ranking” will misallocate budget every quarter.

Limits, maintenance, and when to skip the file

Limitations are part of the deliverable, not a footnote. The file is self-reported. Competitors can copy structure. Major answer engines may never consult it for discovery. Google Search ignores it for ranking and generative visibility per Search Central. Stale links actively mislead the few tools that do expand it into context. Arabic-only audiences gain nothing if linked Arabic pages are weak. Lighthouse agentic checks do not equal SEO scores. Plugin auto-generators that dump every URL recreate sitemap bloat inside a format meant for curation.

Maintenance rules stay simple. Update when the legal name, phone, or primary service URLs change. Validate that every link returns 200. Keep the file under a size a model can skim. Add a one-line note that robots.txt and HTML remain authoritative for access and facts. Re-read Google’s AI optimisation guide quarterly; if Search Central still lists llms.txt among things to ignore for Google Search, keep internal OKRs aligned with that reality.

Skip the file when the site is a thin brochure with no documentation use case, when no tool in the stack reads it, or when leadership only wants it because a competitor announced one on LinkedIn. Shipping emptiness to match a trend still costs review time. Prefer spending that time on the Arabic service page that a procurement officer in Qatar actually opens.

When the file is justified, treat it as documentation ops. Version it in git beside the site. Review it in the same release checklist as sitemap regeneration after a migration. Tell stakeholders in one sentence: optional agent map; zero Google ranking effect; real visibility work remains content, entities, crawl access, and bilingual quality.

Frequently asked questions

Does llms.txt help Google rankings in Qatar?

No. Google Search Central states that Google Search does not use llms.txt or similar AI text files for appearance in Google Search, including generative AI capabilities. The Search Central changelog clarifies that maintaining these files will not negatively or positively impact visibility or rankings. Qatar location does not create a special exception. Ranking work still means crawlable bilingual pages and ordinary SEO fundamentals.

Is llms.txt a ranking signal for AI Overviews or AI Mode?

Google’s generative AI features on Search rely on core Search ranking and quality systems and publicly crawlable content, not on a special llms.txt format. The AI optimisation guide lists LLMS.txt under practices site owners can ignore for Google Search. Claims that the file unlocks AI Overviews for Doha queries contradict that documentation. Focus on indexed HTML that answers the query with unique substance.

Did Google endorse llms.txt because some Google sites publish one?

John Mueller answered that Google properties showing llms.txt is not an endorsement from Search. Different teams may ship documentation helpers for agents without adopting the file as a ranking signal. Presence on `ai.google.dev` or similar hosts is not proof of Search usage. Treat Mueller’s Bluesky clarification as the short answer for stakeholders who equate Google hosting with Google ranking policy.

Should a .qa site still publish llms.txt?

Only if a concrete tool will read it: documentation platforms, coding agents pointed at the domain, or an internal RAG pipeline the company controls. Google Search will neither reward nor punish the file according to published guidance. For many Qatar SMEs without public developer docs, the file is optional vanity unless that tooling need exists. Skipping it to finish Arabic content is often the better trade.

How does robots.txt differ from llms.txt for AI crawlers?

robots.txt is interpreted under the Robots Exclusion Protocol and is checked by compliant crawlers before fetching. It can allow or disallow Googlebot, Bingbot, OAI-SearchBot, Claude-related bots, and tokens such as Google-Extended. llms.txt does not grant permissions and does not replace those rules. Qatar teams that care about training opt-outs or search-bot access must edit robots.txt (and CDN/WAF rules), not Markdown link lists.

What should a bilingual Qatar template include?

Include the public brand name, legal entity if different, Doha or other Qatar location, CR reference as published on-site, +974 contact, and absolute HTTPS links to EN and AR About, Contact, Services, and Privacy pages. State explicitly that the file is optional and does not affect search rankings. Point agents to robots.txt for access rules and to HTML as canonical facts. Keep Optional links short and current.

Can llms.txt replace schema markup or hreflang?

No. Organization or LocalBusiness JSON-LD supports rich-result eligibility and clearer entity parsing on public pages. hreflang pairs English and Arabic URLs for search locale selection. llms.txt cannot perform those jobs, so Qatar teams should keep schema and hreflang on the HTML site and treat llms.txt as an optional map only.