Why AI Doesn’t Recommend Your Brand, and How to Fix It

Short answer: AI assistants such as ChatGPT, Perplexity and Google's AI Overviews build answers from pages they can find and cite, and they lean on what other sources say about a company, not only on what the company says about itself. If your manufacturing business has a good website but little consistent information anywhere else, an assistant has less to work with and may name a competitor instead. The fix is unglamorous: one set of company facts, published the same way on a handful of independent sources that buyers and machines already trust, and linked back to your website.

This guide is for owners of small and mid-sized US manufacturers, job shops, distributors and trading companies. It explains how the major AI search tools say they find and cite pages, what "entity consistency" means in practice, which independent sources are worth your time, and a twelve-week plan you can run with the people you already have. Every rule and number below comes from the publisher's own page, checked on October 21, 2026. Where the official documentation is silent, this article says so instead of guessing.

Why a good website is not enough for AI answers

A strong website makes your company eligible to be cited, but it is still one source talking about itself. When an assistant answers "who makes custom aluminum enclosures in Ohio," it pulls from several pages at once. Companies whose name, location, capabilities and certifications show up consistently across independent sites are easier to confirm and easier to mention.

Google has been explicit about this idea for years, long before AI answers. Its Search Quality Rater Guidelines (version dated September 11, 2025) tell raters in section 2.5 to look for what "outside, independent sources" say about a website, and to trust the independent sources when they disagree with the site's own claims. Section 3.3.1 tells raters to be skeptical of claims that websites make about themselves and to look for independent reviews, references, news articles and ratings instead.

Two honest caveats belong here. First, Google's overview of the rater program says no single rating directly affects how a page or site ranks; the guidelines describe what Google's systems are trying to reward, not a formula. Second, section 3.3.5 of the same document says small businesses often have little reputation information online and that this "is not indicative of high or low quality." Being small is not a penalty. Being invisible beyond your own site simply leaves the machine with less evidence than it has for a competitor who has done the groundwork.

Here is what that looks like in practice. A buyer asks an assistant for "ISO 9001 certified sheet metal fabricators near Columbus that handle prototypes." Your website says all of that. But your LinkedIn page lists an old address, your industrial directory profile still shows the company name you used before a rebrand, and the trade association member list spells your name a third way. A person could sort that out in a minute. A system stitching together facts from many pages has a harder time deciding that these are the same company, and an easier time recommending the fabricator whose details match everywhere.

The common myth is that AI visibility is a website project: add schema, write FAQ pages, publish an llms.txt file, done. Those steps help (we cover them in our AEO technical setup guide), but they all live on the one domain you control. The off-site half of the work is slower and less technical, which is exactly why most of your competitors have not done it yet.

How AI search tools find and cite sources

Each major AI search tool documents how pages get into its answers, and the common thread is simple: your pages must be crawlable and indexable, and the tool then draws on many sources per answer. None of them publishes a formula for which brands get named, so ignore anyone who claims to know one.

Google AI Overviews and AI Mode. Google's AI features documentation says both features may use a "query fan-out" technique, running multiple related searches across subtopics and data sources to build a response. That is why a single answer often cites a mix of sites. Google also states there are no additional technical requirements and no special optimizations needed to appear: a page must be indexed and eligible to show a snippet in regular Search. The same page notes that inclusion is not assured and that recrawling after changes can take anywhere from several days to several months. Traffic from these features is counted in Search Console's Performance report under the "Web" search type.

ChatGPT search. OpenAI's crawler documentation says OAI-SearchBot is the crawler used for search, and that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. GPTBot is a separate crawler for model training, and blocking it does not remove you from search. Robots.txt changes take about 24 hours to be reflected.

Perplexity. Perplexity's bot documentation says PerplexityBot is designed to surface and link websites in its search results and is not used to crawl content for AI foundation models. A separate agent, Perplexity-User, fetches pages when a person asks a question and may include a link to that page in the response.

The table below summarizes what each tool's documentation says you control. The links are in the paragraphs above.

Two practical conclusions follow. One: check that your robots.txt, CDN or firewall is not quietly blocking OAI-SearchBot or PerplexityBot; security plugins sometimes do this by default. Two: because answers draw on several pages, the other pages that describe your company matter. A directory profile, an association listing or a trade article can be one of the pages an assistant reads, even when your own site is not. If you want to see how the tools differ for buyers, our comparison of ChatGPT, Perplexity and Gemini goes deeper.

What entity consistency means for a manufacturer

Entity consistency means every public source describes your company with the same core facts: the same legal name and brand name, address, phone number, website URL, founding year, and the same short description of what you make and for whom. It turns a pile of scattered mentions into one clearly identifiable company.

Search engines call a company an "entity." Machines resolve entities by matching attributes, and mismatches create doubt. Local SEO practitioners have argued for years about how much name, address and phone consistency matters for map rankings. For AI answers the logic is more basic: if two pages disagree about where you are or what you make, a system has to choose, and it may choose neither.

Start with a one-page company fact sheet that everyone in the business uses. Keep it in a shared document and treat it like a controlled drawing: one owner, revision date, no freelancing.

• Legal name exactly as registered, plus the brand or DBA name if different
• Street address formatted one way (Suite vs Ste., abbreviations, ZIP+4 or not)
• Main phone number with area code, and one general inbox
• Canonical website URL (choose www or non-www and https, then never vary it)
• One-sentence description, about 25 to 35 words: what you make, for which industries, from where
• A longer 80 to 120 word description for directories that allow it
• Core capabilities as consistent terms (for example "CNC machining, sheet metal fabrication, powder coating")
• Certifications with certificate numbers and expiry dates, such as ISO 9001 or AS9100
• Identifiers: D-U-N-S number, CAGE code if you sell to government, and any industry registry IDs
• Founding year, employee range and the names of the roles buyers contact (sales, quality)

Then connect the fact sheet to your website through structured data. Google's Organization structured data guide lists properties such as legalName, address, telephone, and sameAs, where you can list multiple URLs of other pages about your organization, such as social or review profiles. It also supports identifiers: Google recommends the iso6523Code field and gives common prefixes, including 0060 for a D-U-N-S number, 0088 for a GLN and 0199 for an LEI. The schema.org definition of sameAs describes it as the URL of a reference page that unambiguously indicates the item's identity.

The sameAs list is where the off-site work and the website meet. Each time you complete a credible profile, add its URL to the list. Each profile, in turn, links back to your canonical website. The result is a closed loop of pages that agree with each other.

A note on Wikipedia and Wikidata, because owners often ask. Wikidata's notability policy accepts items that refer to a clearly identifiable entity that can be described using serious and publicly available references. Most small manufacturers do not yet have those references. Building directory, association and press coverage first is the honest route; an entry created before the references exist is likely to be challenged.

The independent sources that matter for US manufacturers

The best independent sources for a US manufacturer are the ones industrial buyers already use and that verify something about you: industrial directories, business identity registries, professional networks, trade associations, regional manufacturing programs, and trade press. Quality and consistency beat volume; five accurate profiles do more than fifty abandoned ones.

Industrial directories. Thomasnet is the obvious starting point for North American suppliers. Its get-listed page says claiming or listing your company creates a free profile, with an option to upgrade to a paid Verified subscription, and describes a network of more than 500,000 industrial suppliers. Fill in capabilities, certifications and industries served with the exact wording on your fact sheet.

Business identity registries. Dun & Bradstreet's D-U-N-S page states there is no cost to request and receive a D-U-N-S Number, and that most companies receive it within 30 business days; an expedited option costs extra. Many procurement departments ask for this number anyway, and it gives you a stable identifier you can put in your structured data.

Professional network. A complete LinkedIn Page is one of the few profiles nearly every B2B buyer may check. LinkedIn's about page reports more than 1.3 billion members, including more than 290 million in North America. According to LinkedIn's help article on creating a Page, you need a personal LinkedIn account and must confirm you have the right to act on behalf of the company. Use the fact-sheet description in the About section and the same website URL. Our LinkedIn outreach playbook covers what to post once the Page exists.

Trade associations and manufacturing programs. Membership directories of your industry association (for example a precision machining, metal forming or plastics association) are independent confirmation that you are a real company in a specific sector. The same goes for regional programs: NIST's Manufacturing Extension Partnership supports small and medium-sized US manufacturers through state-designated centers, with nearly 1,400 advisors at more than 450 service locations. Projects with an MEP center, state manufacturing associations or local economic development groups often produce member pages, case write-ups or event listings that name your company.

Google Business Profile, if you qualify. Google's Business Profile guidelines allow a profile if your business has a physical location customers can visit or travels to customers where they are. Many plants that receive customer visits for audits or pickups qualify; a warehouse-only address that never receives customers may not. Use your real-world business name with no keywords added.

Trade press and customer stories. A short article in a trade publication about a new capability, a certification or a project, or a case study published by a customer or partner, carries weight that your own blog cannot. Pitch facts, not adjectives: a new 5-axis cell, an expanded cleanroom, a first-article process that cut rework.

This decision table helps you choose where to start. Effort and cost notes reflect the sources linked above; "your time" is the main cost for most items.

Reviews, testimonials and press: what is allowed

Reviews and testimonials can strengthen your third-party footprint, but only if they are real. In the US, the FTC's rule on consumer reviews and testimonials took effect on October 21, 2024, and it bans fake reviews, buying reviews tied to a particular sentiment and undisclosed insider reviews. Ask real customers, and never script or pay for praise.

According to the FTC's questions and answers on the rule, courts can impose civil penalties for knowing violations. The FTC's announcement of the rule lists the main prohibitions: reviews from people who do not exist, including AI-generated ones; reviews from people with no real experience of the business; compensation conditioned on a positive or negative sentiment; insider reviews without clear disclosure; and suppressing negative reviews through threats. The rule is framed around consumer reviews, and the FTC's materials do not single out B2B situations, so the safe practice for a manufacturer is to follow it everywhere.

What does a clean review program look like for an industrial company?

• Ask every customer the same way after delivery, without conditions on what they say
• Get written permission before you publish a customer's name or logo on your site
• Disclose relationships: a distributor partner or an employee's relative must say so
• Reply to critical reviews factually and fix what is fair; do not threaten reviewers
• Keep the original emails or forms behind every testimonial you publish

Press works the same way. A paid placement that looks like editorial coverage is not independent, and buyers notice. Earned coverage is slower, but it is the kind of source reputation research is designed to find.

A twelve-week build plan for a small manufacturer

You can build a credible multi-source footprint in about twelve weeks with a few hours a week from one person: weeks one and two fix your own facts, weeks three to six build the core profiles, and weeks seven to twelve earn third-party mentions and measure. The sequence matters more than the speed.

Weeks 1 and 2: fix the source of truth.

1. Write the company fact sheet and get the owner to sign off on it.
2. Audit what already exists: search your company name, old names, phone number and street address, and list every profile you find.
3. Update your website's About and Contact pages to match the fact sheet word for word.
4. Add or correct Organization structured data with legalName, address, telephone, and identifiers.
5. Check robots.txt and firewall rules so Googlebot, OAI-SearchBot and PerplexityBot can reach your pages.

Weeks 3 to 6: build or correct the core profiles.

1. Create or complete the LinkedIn Page with the fact-sheet description and URL.
2. Claim or create your Thomasnet profile and fill in capabilities, certifications and industries.
3. Request or verify your D-U-N-S Number and correct any outdated details.
4. Create a Google Business Profile only if your location qualifies.
5. Ask your trade associations to update your member listing, and correct old directory entries you cannot delete.
6. Add each finished profile URL to the sameAs list on your website.

Weeks 7 to 12: earn independent mentions.

1. Pick two or three genuine news items: a certification, a new machine, a new market or a solved customer problem.
2. Pitch one to a trade publication or your association newsletter; offer data and photos.
3. Ask two satisfied customers whether you may publish a short case study, and whether they would mention the project themselves.
4. Publish one useful technical article on your own site based on real engineering experience (our guide to turning engineering know-how into authority content shows how).
5. Start a simple AI answer log and repeat it monthly.

The second-order benefit is easy to miss. The same consistency that helps AI tools also helps people. A purchasing manager who checks your LinkedIn Page, your Thomasnet profile and your website and finds the same facts is more likely to send an RFQ. Our article on the trust signals North American buyers look for covers the on-site side of that same checking behavior.

How to check whether it is working

Measure in three layers: whether your facts are consistent across sources, whether AI tools describe you correctly when asked directly, and whether you appear in category answers your buyers would ask. Expect slow movement, because recrawling can take weeks to months, and treat any single answer as a sample, not a verdict.

Layer 1, consistency audit. Once a month, compare every profile against the fact sheet. Count mismatches in name, address, phone, URL and description. The goal is zero, and this is the one metric fully under your control.

Layer 2, direct questions. Ask ChatGPT, Perplexity and Google's AI Mode the same three questions each month, such as "What does [your company] make?" and "Where is [your company] located?" Record whether the answer is correct and which sources are cited. Wrong answers usually trace back to a stale profile you can fix.

Layer 3, category questions. Write five to ten questions a real buyer would ask, like "contract manufacturers for medical device enclosures in the Midwest," and record whether you are mentioned and which competitors are. AI answers vary from one run to the next, so look at trends over several months rather than any single result. Our piece on how North American buyers now use AI to find suppliers suggests how to phrase these questions.

On the analytics side, Google's documentation confirms that clicks from AI Overviews and AI Mode are included in Search Console's Performance report under the "Web" search type, so you will not see them as a separate line. Watch branded search impressions and direct inquiries that mention AI tools. A measurement caveat: set your baseline before you start, or you will not be able to tell whether anything changed.

Where HappyCXO Studio fits

HappyCXO Studio provides a remote team member for small and mid-sized manufacturers and trading companies. For this kind of project, that person can write and maintain your company fact sheet, build and correct the profiles above, keep your website and structured data aligned, and run the monthly AI answer log.

We are based in Taiwan and work remotely with manufacturers and traders, by email and scheduled video calls. Before work starts, we agree on KPIs and a baseline with you, then you get a weekly report and regular reviews. A person reviews everything before it is published in your name, and accounts and profiles stay in your company's name. We do not promise that any AI tool will recommend you; no one can, and the documentation above says as much. What we can do is make sure the evidence about your company is accurate, consistent and easy to find.

If your website itself needs work first, see our website and SEO service, or browse all services. You can also read about our content strategy for AI citations. To talk through your situation, book a 20-minute demo.