Industrial SEO: a guide for industrial companies

The arithmetic that decides which terms are worth having#

A term reporting twenty searches a month at position two produces more requests for quote than one reporting two thousand at position twelve, on every plausible click rate, and the narrow page costs the same day to write. That single comparison decides most of the page list, and it is the opposite of what a keyword volume tool suggests.

If a search for the process you run every day shows competitors, directories, and companies two states away, this arithmetic is usually where the fix starts. It is worth working through once with numbers. The click rates below are assumptions, so substitute your own where you have them.

The narrow term wins, and it keeps winning when you change the assumptions. Traffic arriving on a category word is mixed: students, competitors, people outside your freight radius, and buyers whose part you could never make. Traffic arriving on the exact process in the exact alloy in your state is almost entirely people who could buy.

Why the narrow page is also cheaper

• It takes a day to write, because the facts already exist in your quoting system and your drawings.
• It needs no links to rank, since the competing pages are usually thin or absent.
• It ages slowly. A tolerance range is true for years, so the page keeps working with no maintenance.
• It is the page an assistant can quote, because it contains a specific claim that survives being lifted out.

Building the keyword list from your own quote log#

Your keyword list is already written, in your quote log, your drawings, and your inbox, in the words customers used when they asked. Start there rather than in a volume tool, because the phrases that produce industrial revenue are frequently too small for a tool to report and will therefore never appear in its output.

The usual method is to open a keyword tool, type a category word, and export what comes back. For an industrial company that produces a list of terms held by directories, with the phrases that actually produce your revenue missing entirely, because a tool reporting a minimum of ten searches a month cannot see a phrase searched four times.

The method, in seven steps

1. Export the last hundred quote requests. Pull them from the inbox, the CRM, or the quoting system. You want the words the customer used, which are often different from the words in your catalog. This is the source document for everything that follows.
2. Write down every process, material, and industry named in them. Write each one down exactly as the customer wrote it, such as centerless grinding, 17-4 PH, mechanical zinc, food grade, or ITAR. Keep their wording instead of tidying it into your own, because the words customers use are the words they search with.
3. Add the modifiers that appear alongside them. Look for state names, tolerance figures, standards, quantities, industries, and words like near me, supplier, job shop, and contract. These become the long tail, the many small and specific searches that together produce requests for quote.
4. Check each phrase in a search engine itself. Search it and read the first page. If the results are directories and marketplaces, the term is contested. If they are thin competitor pages or nothing relevant, it is winnable, and a volume tool will probably report no searches for it at all.
5. Group the phrases into one page each. Each process, material, and industry gets a page of its own. If two phrases would be answered by the same page with the same evidence, they belong together. If they need different tolerances or different standards, they belong on separate pages.
6. Rank the pages by gross profit. Multiply the quote requests you expect by your win rate, your average job value, and your margin, and put the highest number first. This usually reverses the order a keyword tool would give you.
7. Write the top five pages before touching anything else. A capability page with real numbers on it takes about a day to write. Five of them is a week of work that changes what the site can rank for, and it is the cheapest week in any search program.

The five modifier families that produce requests

Turning the list into a page map, with URLs#

A good page map gives each process, material, and industry a page at its own address, linked to the other pages it genuinely relates to. A page called Capabilities listing twelve processes ranks for none of them, and that single architectural fault is the most common one on manufacturer websites.

The page map is the deliverable that matters most, and it is usually decided by accident. A site inherits its menu from whoever built the last one, often someone who has long since moved on, and the capability content gets stacked onto one page because that is where it fit in the menu.

A worked map for a plating and finishing shop

Four rules for building the map

• Split when the evidence differs. Two phrases share a page when the same tolerances and standards answer both. They need separate pages when they do not.
• Keep the URL stable and readable. The address should survive a redesign, so build it from the process rather than from the menu label.
• Link across as well as up. A material page should link to every process that applies to it, and each process page back to the material. That is how a reader and a crawler learn what relates to what.
• Give every division its own address when the divisions have different buyers, and keep one site when they share a buyer.

A hub page listing every process and nothing else is doing its job. It is tempting to write four paragraphs of introduction on it, but that turns it into a page competing with its own process pages for the same phrases.

What a capability page has to contain to rank#

A capability page ranks when it contains the numbers a competitor left out. Across sixty-six US manufacturer homepages we measured, four cited a materials or process standard. Publishing your tolerance range, your size envelope, and your standards puts a page ahead of almost everything it competes with, and the facts already exist inside your building." >

There is a template that works, and it is unglamorous. It answers the questions in the order a reader asks them, and it leads with the answer. It also saves your engineers time, because the questions they answer by phone every week are the same ones this page answers in writing.

The structure of a page that ranks and converts

1. The page opens with one sentence saying what you do and for whom, naming the process, the materials, and the region.
2. A specification block comes next. It gives the size envelope, tolerance range, materials and grades, volumes, finish options, and lead time in a table written as text.
3. The standards are named, with their numbers. This is the part that separates you from every other result.
4. The page says what you decline, which filters out the wrong inquiries and reads as confidence.
5. One piece of evidence shows the work. That can be a part you made, the problem it solved, and the tolerance you held.
6. The quote path follows, with a file upload and a named person.
7. The page ends with links to the related process, material, and industry pages.

What changes when the answer is written for the reader#

Google's position is that good SEO is good GEO, the work of being cited in AI answers: its generative features run on the same ranking systems and retrieve from the same index, so the work that earns a ranking largely earns a citation. What changes is the writing. A sentence carrying a process, a number, and a qualifier can be lifted whole and attributed. An adjective cannot. What also changes is measurement. Search Console has begun reporting impressions from AI features, and a click from an assistant still arrives with no referrer, so your quote form remains the only place that click is counted.

The practical advice is the same as everything above. Publish specifics as text, keep the technical layer clean, and make sure search engines can tie your company to a place. Google's own VP of Search and Commerce Global Ads Solutions puts the same point more bluntly: good SEO is good GEO, because the generative features are built on the same core ranking systems and retrieve from the same index, in a signed piece by its VP of Search and Commerce Global Ads Solutions.

Two of the checks in this guide are quicker to run than to read about. A scored pass over your own quote path grades what happens after somebody decides to ask, and the profile scorecard covers the one surface that returns the most for the least work.

Three things you can take off the list

If an agency has pitched you something vague with AI in the name, it was probably one of these. Each is being sold to manufacturers right now as necessary work for the AI era, and Google has stated on the record that Search requires none of them. Ten minutes reading the source can save you a retainer.

• Do not publish an llms.txt file and expect Google to read it. Google has said llms.txt and similar files are not needed for Search. The file costs nothing to publish and should carry no part of a plan.
• Do not buy a tool because it scores your AI visibility. Google does not evaluate third-party SEO tools, and those tools hold no access to its internal metrics. A score from one of them is that vendor's model of Google, sold as a reading of Google.
• Do not commission a separate GEO retainer alongside an SEO one. Ask any agency proposing the second one to name the work in it that is absent from the first. The real answer is writing and measurement, which is one scope of work with two names on the invoice.

One caveat about the source is that it is Google writing about Google, by an advertising executive and not a search engineer, so read it as a statement of what Google says it does.

The three writing habits that matter

• Put a complete answer near the top of the page, in one or two sentences that stay true when quoted out of context.
• Attach a number to every claim you want repeated. That can be a tolerance, a temperature, an hour count, or a standard.
• Give each question its own address. A question as a URL is the exact shape an answer engine quotes.

Measuring search when the sales cycle runs nine months#

Count requests for quote by query rather than sessions. Traffic that never asks for a price is a cost, and a search program judged on sessions gets defunded at exactly the wrong moment, because rankings move months before revenue does. Three imperfect measures used together beat any single number.

The measurement problem in industrial search is that the useful outcome arrives late, and often by phone. A page can rank in month three, produce a request in month five, and bring a purchase order in month eleven, by which point the connection between them is invisible to any analytics tool.

Set the reporting rhythm to match the cycle. Look at impressions and requests every month, at whether the requests are worth quoting every quarter, and at revenue attributable to search once a year. Reviewing revenue monthly on a nine month cycle produces a decision to stop, taken four months before the program would have worked.

What actually happens in months one to twelve#

Months one and two produce pages and no visible result. Months three to six produce impressions and the first requests on narrow terms. Months six to twelve are when compounding becomes obvious. Nothing about this is fast, and any program judged at week six will be judged before it has produced anything.

Two things change this timeline materially. A site that already has authority and traffic sees movement faster, sometimes within weeks, because the new pages inherit standing. A site that has just been rebuilt and lost its redirects will go backwards for a quarter before it goes forwards, which is the strongest practical argument for mapping every URL before a relaunch.

The faults we find most often on industrial sites#

Eight faults account for most of what we find on industrial sites, and seven of them are free to fix. The most expensive is the one that looks like progress: buying traffic to a site that cannot answer a tolerance question.

1. One page covers every capability. It ranks for none of them, and splitting it is the single highest-return change available.
2. The copy uses category language in place of specifications. Precision manufacturing solutions matches nothing anyone types.
3. Specifications live only in PDFs. The best content on the site is doing no work.
4. Certifications are shown as logos. An image carries no text, no scope, and no registrar.
5. The company chases the head term. A year goes into a phrase held by directories, while the winnable pages stay unwritten.
6. Traffic is bought before the site can answer. Paid clicks landing on a page with no tolerance figure leave without asking for a price.
7. Redirects are skipped at relaunch. It is the most reliable way to lose every ranking at once.
8. Search is judged on sessions. The number moves for reasons unrelated to revenue, in both directions.

Questions industrial companies ask about search#

Here are eight questions industrial companies ask about search, each answered so it can be read on its own.