--- title: "Local Government Website Search: Residents and AI - Natalia" description: "Why search on a local authority website disappoints, why the meta keywords tag does not help Google, and how to move to questions asked as sentences." url: https://getnatalia.com/en/local-government/website-search lang: en --- Local government # Searching a local authority website: from “What are you looking for?” to “What is your question?” The search box expects a keyword. Residents arrive with a question, and they have got used to asking it to an AI. What that changes for a municipality's website. [Request a demo](https://getnatalia.com/en/contact) 48% of people aged 12+ in France use generative AI (Arcep, 2025 Barometer) 73% of those users rely on it to look for information 2009 year Google announced it ignores the meta keywords tag €43 per month excl. VAT, our chat on an annual plan ## Table of Contents 1. [Why on-site search disappoints residents](https://getnatalia.com/en/local-government/website-search#section-0) 2. [A growing share of residents already use AI to look things up](https://getnatalia.com/en/local-government/website-search#section-1) 3. [Hand-typed keywords do not help Google ranking](https://getnatalia.com/en/local-government/website-search#section-2) 4. [Moving to “What is your question?”](https://getnatalia.com/en/local-government/website-search#section-3) 5. [The website stays the source of truth](https://getnatalia.com/en/local-government/website-search#section-4) 6. [Where to start](https://getnatalia.com/en/local-government/website-search#section-5) ## Why on-site search disappoints residents A local authority's website search compares words. A resident types “canteen registration”, the engine looks for those words in the pages, and last year's news item may come up before the procedure page. When the engine can't even find the articles the authority published, the resident may give up or call the front desk. The staff who publish are not to blame. A keyword engine depends heavily on the page's vocabulary, and residents use their own words: “canteen” rather than “school catering”, “tip” rather than “recycling centre”. ## A growing share of residents already use AI to look things up According to the Digital Barometer published by Arcep, the French telecoms regulator, in February 2026, 48% of people aged 12 and over in France use generative AI services, up fifteen points in a year. Among those users, 73% use them to look for information. These residents are used to typing a full question and getting a written answer. On the website's search box, they find the old way of working. And when they ask a general-purpose AI, the answer does not necessarily come from the authority's own website. ## Hand-typed keywords do not help Google ranking Many teams filled in keyword lists for every page, thinking it would improve their search ranking. Google's developer documentation states that the meta keywords tag is not used by Google Search and has no effect on indexing or ranking. Google had already announced this in 2009. That time is better spent on a clear title, a page that uses residents' words and up-to-date information. Those are also the pages an assistant answers best from. ## Moving to “What is your question?” A conversational assistant reads the website pages and published documents, understands a question asked in everyday language and answers by citing the page it used. A resident who asks which documents are needed to sign their son up for the canteen gets the list and a link to the procedure. The existing search box can stay. With us, the assistant is added to most websites with one line of code and takes over as soon as the question is asked as a sentence. [See our agent for resident enquiries](https://getnatalia.com/en/local-government) ## The website stays the source of truth An assistant that answers from the website makes every page more useful, as long as it is accurate. An outdated page gives an outdated answer. Publishing is still necessary. The questions asked also show what residents are looking for. With the statistics in our Pro plan, recurring topics show what comes up most, which helps you decide which pages to complete first. [How a chatbot cites its sources and stays within its scope](https://getnatalia.com/en/local-government/sourced-answer-chatbot) ## Where to start List the ten questions your front desk gets most often, by phone, in person and by email. Check that each one has an up-to-date page on the website. Then choose the pages and public documents the assistant may read, and test it on those ten questions before opening it to residents. With us, there is no setup fee, and the Essential plan covers 500 conversations and 500 indexed pages for €43 excl. VAT per month on an annual plan. [See the full pricing grid](https://getnatalia.com/en/pricing) ## Keyword search or an assistant that answers | Criterion | On-site search | General-purpose AI, outside the website | Assistant on the website | | --- | --- | --- | --- | | What the resident types | One or two keywords | A full question | A full question | | What they get | A list of pages | A written answer | A written answer and the source page | | Sources | The website pages | Not controlled by the authority | Pages and documents chosen by the authority | | When information is missing | No result, or an off-topic one | Depends on the tool | The assistant is set up to say so and pass the request on | | Updates | Depends on the page's wording | Depends on the tool | Based on the published pages | Sources: Google Search Central, meta tags documentation (updated 10 December 2025); Arcep, Digital Barometer 2025 (press release of 9 February 2026). ## Sources - [Google Search Central: meta tags and attributes that Google supports](https://developers.google.com/search/docs/crawling-indexing/special-tags), accessed 1 October 2026 - [Google Search Central Blog: “Google does not use the keywords meta tag in web ranking”](https://developers.google.com/search/blog/2009/09/google-does-not-use-keywords-meta-tag), September 2009 - [Arcep: Digital Barometer 2025, equipment and usage (in French)](https://www.arcep.fr/actualites/actualites-et-communiques/detail/n/equipements-et-usages-du-numerique-090226.html), press release of 9 February 2026, Credoc survey of June 2025 ## Frequently asked questions ### Why doesn't our website search find our own articles? Because a keyword engine compares words. If the resident doesn't use exactly the page's vocabulary, or if a news item contains the same words, the right page drops down or doesn't appear. ### Does the meta keywords tag still help SEO? Not with Google. Its documentation states that the meta keywords tag is not used by Google Search and has no effect on indexing or ranking. ### Do residents really use AI to look for information? According to Arcep's Digital Barometer published in February 2026, 48% of people aged 12 and over in France use generative AI services, and 73% of those users rely on them to look for information. ### Should we remove the website's search engine? Not necessarily. The assistant is added to most websites with one line of code, and the search box can stay for those who use it. ### Can an assistant answer from published PDFs? Yes. With us, a file counts for its number of pages in the indexed volume. A 20-page set of rules therefore uses 20 of the 500 pages in the Essential plan. ## Go further - ### Chatbot for local government One agent for resident enquiries, on the website and by email - ### A chatbot that cites its sources Scope, non-public content, handover to staff - ### Custom chatbot The chatbot trained on your documents ## See the assistant answer on your pages A 30-minute demo, based on your website. [Request a demo](https://getnatalia.com/en/contact)