Original Information Gain: The Local Business SEO Advantage

Original information gain is the new, useful knowledge a page adds beyond what a searcher can already find. For a local business, it usually comes from counting something real in your own enquiries, quotes or jobs, publishing the method and limits honestly, and saying what you changed as a result. It needs discipline, not a research budget.

The internet does not need another article saying fast replies are important. A local business can publish something more valuable: what its customers actually ask, which details make a lead actionable and where its own process breaks down.

What original information looks like#

You already hold data no competitor can copy. It is in call notes, enquiry forms, quote emails and job records. A few examples of what that could become:

• A plumber anonymises and classifies 100 enquiries by service, suburb, urgency, time received and missing information.
• An electrician publishes the ten questions that most often require a site inspection, and why.
• A cleaner compares the details needed for end-of-lease, recurring and commercial quotes.
• A clinic explains which booking questions an assistant may answer and which must go to trained staff.

Each of these answers a question customers have, in a way only that business can.

Use an evidence ladder#

Not every claim carries the same weight. Label the rung honestly:

1. Business opinion. "We think most people want a callback, not an email."
2. Observed pattern, no count. "We often see enquiries without a suburb."
3. Counted internal sample, with method and period. "Of the enquiries we received in a stated month, this many had no suburb."
4. Repeatable test or benchmark. The same count, run again after a change, using the same definition.
5. Independent validation. Someone outside the business checks or reproduces it.

A hypothetical example of rung three: "In 43 anonymised website enquiries received during September, 12 did not include a suburb." That is stronger than "customers always forget their suburb", but it is still one business's sample, not a national statistic. Say so on the page.

Publish the method with the finding#

Original content earns trust when readers can see how it was produced. For any counted finding, include:

• The period the data covers.
• The sample: how many enquiries, jobs or conversations, and from which channels.
• The definition: what counted as "missing a suburb" or "urgent".
• The limits: seasonality, one location, a change in advertising during the period.
• The action: what you changed, and when you will count again.

Remove personal details before anything leaves your systems. Do not select only flattering results. If the data shows your chatbot failed to collect a suburb in several conversations, say what you changed and when you will repeat the test. That candour is the information gain.

Why this matters for search and AI answers#

Google's guidance on generative AI features emphasises unique, non-commodity content and advises prioritising helpful, original material over shortcuts (optimising for generative AI features; new controls and insights for website owners). Its guidance on generated content also says that using automation to produce many pages without adding value for users may breach its spam policy on scaled content abuse (using generative AI content).

Academic work on generative engine optimisation suggests that adding cited sources and statistics can improve a source's visibility in generated answers in experimental settings (GEO: Generative Engine Optimization). Those are experimental findings across particular engines and query sets, not a formula for any individual site, and quality and relevance still matter. A real, counted, honestly limited finding from your own business is a better input than an invented statistic.

A simple process to start#

1. Pick one question customers ask often, or one detail your team often has to chase.
2. Choose a period, such as last month's website enquiries.
3. Anonymise the records: remove names, numbers, addresses and anything sensitive.
4. Count one thing with a written definition.
5. Write it up with the method, limits and what you will change.
6. Repeat after the change, with the same definition.

Where to find the material depends on your tools. If you use Pivra on Starter or above, the Leads page shows captured enquiries with a transcript link and can be exported as CSV, and conversations can be exported from the Conversations tab (lead capture; analytics). Treat those exports as personal information: limit who handles them and strip identifying details before analysis.

What not to do#

• Do not invent numbers, including "illustrative" ones presented as findings.
• Do not round a small sample into a sweeping claim.
• Do not publish customer details, quotes or stories without permission.
• Do not chase volume. One counted finding is worth more than ten generic posts.

For practical page structures that put this kind of evidence to work, see the answer page template.

Next step: count one real customer pattern this week, and publish the method alongside the finding.

Sources#

• Optimising your website for generative AI features on Google Search
• New opportunities, control and insights for website owners, Google
• Google Search's guidance on using generative AI content
• GEO: Generative Engine Optimization (arXiv)