Now accepting new projects — Get a free prototype →

Editorial

How we decide what to publish.

We publish a lot about web design, local SEO and site performance. Most of that subject matter is crowded with confident advice that is out of date, copied from someone else's copy, or quietly selling something.

These are the rules we hold ourselves to so you can tell the difference. They are specific on purpose — a guidelines page that only promises to be “accurate and helpful” commits to nothing.

Sourcing

  1. Primary sources, always. A claim about Core Web Vitals cites Google. A claim about accessibility cites the W3C. A claim about how DNS behaves cites the RFC. We do not cite a blog citing a press release citing a study, and we link the source so you can check it rather than trust us. The glossary is the clearest example of the standard.
  2. Numbers carry a date. Performance thresholds move — Google has changed the Core Web Vitals set twice, retiring First Input Delay in March 2024. Any figure we publish is true as of a stated review date, and pages carry that date at the bottom.
  3. We say when something is our opinion. “Google treats LCP under 2.5 seconds as good” is a fact with a source. “Chasing a perfect Lighthouse score is a poor use of your budget” is our judgment. We do not blur the two together to make an argument sound more settled than it is.
  4. We do not publish a statistic we cannot trace. The round numbers that circulate in this industry — conversion lifts, bounce-rate cliffs, “53% of users abandon after 3 seconds” — are usually traceable to a single vendor study with a commercial interest, if they are traceable at all. If we cannot find the original, we leave it out.

How AI is used here

We use AI in our work and we are not going to be coy about it. The full picture, including the chat widget and our internal tooling, is on the AI disclosure page. For published writing specifically:

  1. AI drafts. A human decides. Articles are often drafted with AI assistance from an outline and a set of sources. Nothing publishes on the strength of the model being confident. A person is responsible for every claim that goes live, and that person is named on the byline.
  2. Facts get checked against the source, not the draft. The failure mode of AI drafting is not bad prose, it is a plausible number with no origin. Every figure and threshold is verified against the primary source before publication, because a model will produce a citation-shaped sentence whether or not the citation exists.
  3. The byline is a person, not a brand. Everything is published under a named author with a real profile, not an anonymous “staff” byline. If a piece is wrong, you know exactly who to hold to it.

Every post has been through this

All 187 articles in the archive have been through the editing pass described above, not just the recent ones. We track it per post rather than declaring it done, because “we review everything” is easy to say and hard to check.

If you find a post that does not meet what is written here, that is a mistake rather than a policy. Tell us, and it gets fixed and logged on the corrections page.

Independence and money

  1. We sell websites, and that is the bias to watch for. Wildcore makes money building and maintaining sites for local businesses. That gives us an obvious interest in you concluding that your website matters. Where our advice happens to point at hiring someone, assume the conflict and weigh it accordingly.
  2. We will tell you when the answer is not a website. For a lot of local service businesses, a claimed Google Business Profile drives more calls than a new site would, and it is free. We say so in the glossary and we say so on calls, including when it costs us the job.
  3. No paid placements, and affiliate links are labelled. We do not accept payment to recommend a tool or to feature a company in an article. Where a link earns us anything, it is disclosed at the point of the link, not in a footer nobody reads.

Corrections

When we publish something wrong, we fix the page and log it publicly, with what it said before, what it says now, and why it was wrong. The log is append-only — nothing gets quietly removed once it is on the page.

Silent edits are the thing that makes a publication untrustworthy, because they make it impossible to know whether anything you read was ever different.

What we will not do

  • Publish fake reviews, fake ratings, or testimonials from people who did not say them.
  • Mark up content in schema that a visitor cannot see on the page. That is the line that turns structured data into a manual penalty, and it is the most common way small business sites get burned by an SEO contractor.
  • Republish another site's article with the words shuffled, or generate near-identical pages for a list of towns purely to occupy search results.
  • Give legal, medical, financial or tax advice. Where accessibility law comes up, we describe the technical standard and say plainly that it is not legal advice.

The standards we borrow from

None of this is invented. The sourcing and corrections rules follow established practice from organisations whose whole job is publishing credibility:

Last reviewed July 31, 2026. We date these pages by review, not by deploy — a rebuild does not change when a human last read this.