Google Published Guidance on Hiring an SEO. Read It.

Most objections an independent SEO meets are the same question in different words: how does a business owner tell good advice from confident advice? Until this summer there was no neutral document to point at. On 5 June 2026 Google published guidance on third-party SEO tools, services and advice, and rewrote its long-standing page on whether you need an SEO at all. Both pages carry a last-updated stamp of 2026-06-05 UTC, and between them they are the most direct thing Google has written about how to evaluate the people selling you search work.

What the two pages are, and what they are not

Neither page is a directory, a certification scheme or an endorsement of anybody. The guidance on third-party SEO tools, services and advice is a framework for reading claims. Its position is that helpful advice either presents itself as opinion or cites official documentation, while the unhelpful kind misrepresents Google. Google states directly that it doesn’t evaluate third-party services, so be wary of such claims and those making them, which disposes of every badge, partnership logo and “Google-approved” line in a cold email in one sentence.

The companion page is more practical. It lists what a competent SEO does: reviewing site content and structure, technical advice on development, content development, managing online business development campaigns, keyword research, training, market and geography expertise, and — newly — optimising for generative AI. That last item is quietly significant given how much argument there has been about whether generative engine optimization is a real discipline. Google has put it on a list of ordinary SEO services rather than treating it as a separate trade.

The claims Google singles out

The warning signs across the two pages cluster tightly, and they are worth reading as a checklist before you sign anything:

• Unsolicited outreach email offering SEO services.
• Secrecy about what will be changed, or refusal to explain a recommendation.
• A request for write access to Search Console at the start of a relationship.
• Claims of a special relationship with Google, or of priority submission.
• Guaranteed rankings. Google’s own wording is that no one can guarantee a #1 ranking on Google.
• Advice that conflicts with Google’s published documentation.
• Requests for large numbers of links, or bulk submission to directories.

None of that is new to anyone who has worked in the field for a decade. What is new is that it now sits on developers.google.com with a date on it, which turns one practitioner’s opinion against another’s into a documented reference either party can check.

An audit score is not a ranking signal

The sharpest line in the third-party guidance is about data. Google states that third-party tools don’t have access to our internal ranking data, and that using a tool or service does not guarantee ranking success. That is worth sitting with, because the audit score is the most common artefact in SEO sales. A tool that reports a site at 62 out of 100 is reporting its own opinion of its own checklist. The checklist may be well built and the recommendations sound, but the number measures nothing Google measures, and improving it is not the same as improving anything.

The same reasoning applies to the evidence on AI content detectors: an output with a confident number attached is not evidence of the thing it claims to detect. The tools worth paying for surface a fact you would otherwise find by hand — a broken canonical, a 404 in a sitemap, a template shipping the wrong markup. The score on top is packaging.

The questions the documentation tells you to ask

The list of questions on Google’s page on whether you need an SEO is short, and each item maps to something checkable. Ask for examples of previous work. Ask whether the work follows Google Search Essentials. Ask what results to expect, how they will be measured and over what period. Ask about experience in your market and geography. Ask how changes will be reported. Ask for references.

The measurement question separates most pitches. An engagement that cannot state its baseline before work starts cannot demonstrate anything afterwards, and “we will watch the rankings” is not a baseline. Search Console impressions, clicks and average position, segmented by page group and country and recorded on the day the contract starts, is.

What this looks like from the supplier side

Reading these pages as a supplier rather than a buyer is uncomfortable in a useful way. Working at Rank Math puts a steady stream of support questions in front of me that begin with something an agency told a site owner, and a fair proportion of those beliefs trace back to a tool’s output being read as Google’s opinion. Documented changes, a stated baseline, an explanation for every recommendation and no guarantees at all is the defensible shape — and a claim like what llms.txt actually does tests whether a supplier is citing evidence or repeating a convention. If the answer to “why are we doing this” is “everyone does”, the engagement is already off Google’s own map.