Field notes 001 · Our own website
Our SEO experiment.
The work starts here.
An SEO agency should make its own work inspectable. This is FlatRelay’s website audit: what we observed on Google, what needs to improve, and how we will evaluate the outcome.
Baseline report · Growth results not yet available
The brand was discoverable, but two inspected commercial searches did not surface FlatRelay on page one. We found concrete trust and page-quality gaps. This report records a starting point; it does not claim increased traffic, leads, or revenue.
Observed, not estimated
Our SEO experiment starts with a baseline.
Manual browser checks used English Google results with US targeting and personalization disabled. The browser did not establish a precise US location. These observations are snapshots, not a continuous rank-tracking study.
Flatrelay
Homepage first among organic results; a branded AI Overview described the agency.
One branded search observation, not evidence of new-customer acquisition.
site:flatrelay.com
Seven commercial pages appeared in the inspected results.
A site search is not a complete count of indexed pages.
SEO for manufacturers
FlatRelay was not on the first results page inspected.
One query and one observation; no claim about its exact position beyond that page.
manufacturing SEO agency
FlatRelay was not on the first results page inspected.
Results can vary by location, device, and time.
Evidence source: browser observations of Google and the public FlatRelay site on September 9, 2026. Raw screenshots and exports are not attached to this initial report. These are self-reported observations, not independent verification.
Evidence → action → verification
Five things worth fixing first.
These findings explain the changes in this website revision. Implementation checks establish that a change works; only later measurement can establish whether search performance improves.
01 / Trust
A demo score looked like live evidence.
- Evidence
- The homepage displayed 72/100 with a live label, while the source used a fixed example value.
- Action
- Label the score as illustrative and link to this dated audit.
- Acceptance check
- A visitor can distinguish an example from an actual measurement.
02 / Quality
Commercial pages contained unfinished text.
- Evidence
- The About page included an unfinished founder notice. The audit page rendered object values instead of AI engine names.
- Action
- Remove unfinished text, use the verified founder name, and render each engine’s label.
- Acceptance check
- Review generated pages for publishing placeholders and broken values.
03 / Relevance
The offer did not clearly explain Google SEO.
- Evidence
- The homepage and manufacturer page led with AI visibility. The inspected buyer searches included manufacturing SEO intent.
- Action
- Explain technical SEO, capability content, and the path from search to a qualified RFQ.
- Acceptance check
- The title, introduction, service scope, and next step describe the same offer.
04 / Technical
Canonical URLs need one consistent format.
- Evidence
- The site exposed both slash and non-slash URL variants while canonical tags used non-slash URLs.
- Action
- Use non-trailing-slash URLs consistently in canonical tags, internal links, and the sitemap, with redirects from slash variants.
- Acceptance check
- Each generated page’s canonical matches its intended public URL.
05 / Buying confidence
The sample did not show an actual audit.
- Evidence
- The sample report section showed an illustrative diagram rather than a detailed record of findings.
- Action
- Make this self-audit available without a form and connect it to service scope and pricing.
- Acceptance check
- A buyer can inspect evidence, proposed fixes, and limitations before making an inquiry.
The next evidence
Measure buyers, not just positions.
Google search and inquiries
Establish Search Console access, filter to United States, and separate branded from non-branded queries. Record clicks, impressions, CTR, and landing pages. Connect successful inquiries to qualified conversations, proposals, and revenue where attribution is available.
Current status: traffic and conversion baseline unavailable.
AI citation testing
Agree a fixed set of buyer prompts. Record engine, model where available, date, location, repetitions, mentions, citations, and factual errors. Keep this dataset separate from Google click and lead reporting.
Current status: no multi-engine baseline published.
A reusable progress log
For each release, record the page, date, problem, change, and verification. After enough data accumulates, compare equivalent 28-day periods and show absolute counts alongside percentages. Annotate other marketing activity and keep unsuccessful changes visible. A before-and-after comparison alone does not prove causation.
What this report can and cannot prove.
This is a self-audit of an agency website, not a manufacturing client success story. The process can inform a supplier audit; its results cannot be transferred to a client or used to predict a ranking.
PageSpeed Insights showed no field data during the audit and did not return usable lab measurements. Automated HTTP requests returned 403 responses while normal browser pages loaded. We therefore did not conclude that Googlebot was blocked or that the site passed Core Web Vitals.
Google says SEO fundamentals remain relevant to AI Overviews and AI Mode, with no special AI files or schema required. We treat llms.txt as an optional experiment, not a prerequisite or a proven ranking factor. Read Google’s guidance on AI features.
The FlatRelay AI Visibility Score is our diagnostic rubric. It is not a Google metric, a probability of being recommended, or a revenue forecast. We publish the underlying evidence and limitations alongside any score we report.
See what this approach means for your site.
For manufacturers serving US buyers, we start with your capabilities, buyer questions, and proof. Review the service and exact scope before committing to an audit.